WO2005059771A1 - 対訳判断装置、方法及びプログラム - Google Patents
対訳判断装置、方法及びプログラム Download PDFInfo
- Publication number
- WO2005059771A1 WO2005059771A1 PCT/JP2004/015263 JP2004015263W WO2005059771A1 WO 2005059771 A1 WO2005059771 A1 WO 2005059771A1 JP 2004015263 W JP2004015263 W JP 2004015263W WO 2005059771 A1 WO2005059771 A1 WO 2005059771A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- sentence
- translation
- natural
- words
- word
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/42—Data-driven translation
- G06F40/45—Example-based machine translation; Alignment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/42—Data-driven translation
- G06F40/47—Machine-assisted translation, e.g. using translation memory
Definitions
- the present invention relates to a bilingual translation judging device, a method, and a program, and more particularly, to a bilingual judging device for judging at least a translation target phrase in an original sentence of a source language, and a bilingual judging applicable to the bilingual judging device.
- the present invention relates to a method and a program for causing a computer to function as the translation judging device.
- a so-called machine that uses a computer to translate a sentence (original text) written in one natural language (source language) into a text (translation sentence) written in another natural language (target language).
- source language natural language
- translation sentence text written in another natural language
- Patent Document 1 a set of expression forms represented by collocations is stored in an English-Japanese collocation dictionary prepared in an HD device, and in a syntactic analysis process, words combined by coordination conjunctions are used.
- the composed expression form is searched in the English text and the searched expression form is stored in the English-Japanese collocation dictionary, or when the words constituting the searched expression form have the same prefix or suffix.
- Patent Document 2 a feature table in which the meaning and the feature are associated with each polysemy is stored in advance, and a feature record is generated for the input original language of the first language, and the generated feature table is generated.
- a technology that compares a record with a feature table and selects and outputs the meaning of a polysemy based on the features of the original sentence.
- Patent Document 1 JP-A-11-328178
- Patent Document 2 JP-A-6-314294
- the work of translating an original sentence written in a source language into a translated sentence written in a target language is generally performed.
- words in the source language written in the source language are converted to words in the target language, and the converted words are rearranged according to the grammar rules of the target language. It goes through a translation process.
- This translation process is also used in machine translation, which translates using a computer.Registration is registered in the dictionary in units of words, and the original literacy is also extracted in order, and the translation of the extracted words is searched.
- the original sentence is replaced by a word-by-word translation, and the part of speech of each word in the original sentence is determined and the syntax is analyzed. It is common to obtain translations (translations) by rearranging the translations.
- the present invention has been made in view of the above fact, and it is highly possible that a natural bilingual sentence can be obtained as a target language sentence from a source language original sentence. It is an object of the present invention to obtain a translation judging device, a translation judging method and a program capable of obtaining a proper translation.
- a bilingual determination device includes a storage unit configured to store a plurality of natural sentences of a source language including a plurality of words in association with a bilingual sentence of a target language, Search means for searching a plurality of natural sentences of the source language stored in the storage means for a natural sentence including a translation target phrase in the original language of the source language, and extracting by a search by the search means The degree of coincidence between the obtained natural sentence and the original sentence is obtained, and based on the obtained degree of coincidence, at least the translation target phrase in the translated sentence of the natural sentence selected is converted into at least the translation target phrase in the original sentence And first translation judging means for judging that the translation is a translation.
- a plurality of natural sentences in the source language composed of a plurality of words are stored in the storage unit in association with the translated sentences in the target language.
- the natural sentence according to the present invention is a sentence, a phrase, a collocation, a fixed expression, a collocation, etc., which is not edited and processed such as division in words or extraction of polysemous words, such as a dictionary in conventional machine translation.
- a translation corresponding to a natural sentence may be used as a natural language sentence.
- words and their translations are also used. It may be stored in the storage means.
- the source sentence of the source language (any one of a sentence, a phrase, a collocation, a fixed expression, and a collocation) may be used.
- the natural sentence including the translation target phrase in (i) is searched by the search means.
- the word to be translated is a word to be translated in the original text, and may be a word or may be composed of a plurality of words.
- the translation judging device according to the present invention is used as an electronic dictionary when a human (translator) translates, the translation target phrase is specified by the translator.
- the translation target phrase performs machine translation using the bilingual determined by the bilingual determination device according to the present invention.
- the bilingual determination device Specified by machine translation equipment or automatic translation equipment.
- the user may be allowed to specify the original sentence that contains the translation target phrase, or it will be automatically determined (for example, the sentence or phrase containing the translation target phrase is automatically identified as the original sentence). Judgment, etc.). Since the search means searches for a natural sentence that includes the phrase to be translated, this search extracts a natural sentence that includes the translation of the word to be translated in the corresponding bilingual sentence. .
- the first bilingual judging means obtains a degree of coincidence between the natural sentence extracted by the retrieval means and the original sentence, and selects based on the obtained degree of coincidence. At least the translation of the target phrase in the translated sentence of the natural sentence is determined to be a translation of at least the target phrase in the original sentence.
- the natural sentences for example, natural sentences having the same meaning using polysemous words existing in the original sentence
- the natural sentences are close to the original sentence, and have a high probability and a high probability. Is selected as a natural sentence, and a natural bilingual sentence corresponding to the selected natural sentence is obtained.
- At least a bilingual translation of the translation target phrase in the bilingual sentence is a translation of the translation target phrase in the original sentence.
- a bilingual translation of a phrase other than the translation target phrase in the bilingual translation may be determined to be a bilingual translation of the phrase in the original text, needless to say, ).
- the invention described in claim 1 converts a natural sentence in the original language into a bilingual sentence in the target language.
- a natural sentence in the original language are stored in memory, and stored in the source sentence.
- a high natural sentence is selected, and a translation in a natural bilingual sentence corresponding to the selected natural sentence is determined to be at least a translation of the target phrase in the original sentence.
- the invention described in claim 1 can be realized by storing a natural sentence and a bilingual sentence in the storage means. Therefore, at least when the natural sentence and the bilingual sentence are stored in the storage means, the natural sentence is stored in word units. This eliminates the need to perform complicated editing and processing such as dividing by, extracting ambiguous words, listing all possible translations of ambiguous words and associating them. Further, in the invention described in claim 1, it is possible to obtain a match between the natural sentence extracted by the search means and the original sentence, and select a natural sentence based on the obtained matching degree to obtain an appropriate translation. Therefore, there is no need to perform complicated processing such as part-of-speech determination and syntax analysis in conventional machine translation, and the processing can be simplified.
- the retrieval unit matches the original sentence completely from a plurality of natural sentences of the source language stored in the storage unit.
- the first bilingual judging means searches the bilingual sentence of the perfectly matched natural sentence with the translated sentence of the original sentence. It is preferable to make a judgment. Thereby, when a natural sentence that completely matches the original sentence is stored in the storage means, a bilingual sentence of the original sentence can be obtained.
- the matching degree between the natural sentence extracted by the search and the original sentence can be obtained, for example, as follows. That is, in the invention of claim 3, in the invention of claim 1, the first translation judging means counts and counts the number of matching words between the natural sentence extracted by the search means and the original sentence. The feature is to evaluate the degree of coincidence with the original text so that the degree of coincidence with the original text increases as the number of matching words increases. The number of matching words is an important index indicating the degree of matching between the natural sentence and the original sentence. By evaluating the matching degree so that the matching degree increases as the number of matching words increases, the A natural sentence close to the original sentence can be selected with high accuracy based on the criticality.
- the degree of coincidence for example, the counted number of matching words can be used as it is. It is preferable to obtain a value obtained by dividing the number of words by the number of words constituting the translation target phrase, and use the obtained value as the degree of coincidence.
- the matching score according to the present invention is a value obtained by normalizing the number of matching words based on the number of words forming the translation target phrase. Therefore, by using this matching score, the word forming the translation target phrase is obtained. Regardless of the number, natural sentences close to the original can be selected with higher accuracy.
- the matching word is a word constituting the translation target phrase or another word, and the number of matching words of the words constituting the translation target phrase and the matching of the other words are determined.
- a value obtained by multiplying the number of words by a different weight (a weight set so that the words constituting the translation target phrase have a higher weight) (evaluation value of the number of matching words) may be used as the number of matching words. ! ⁇ .
- the first bilingual judging means also counts the number of mismatched words between the natural sentence extracted by the search means and the original sentence, as described in claim 5. Then, the degree of coincidence with the original sentence may be evaluated so that the degree of coincidence with the original sentence increases as the counted number of unmatched words decreases.
- the number of unmatched words is also an important index indicating the degree of matching between the natural sentence and the original sentence, along with the number of matched words, and the number of unmatched words is small using the above-mentioned unmatched words in addition to the number of matched words described in claim 3.
- a natural sentence is stored in the storage means. Therefore, in the invention of claim 3, for example, "a”, “the”, “to”, “in” or the like in English sentence. If words that appear frequently in the natural language of the source language are determined to be matching words, the source text that contains many of these frequently occurring words will not be close to the original text due to the effects of the frequent words. A sentence may be incorrectly selected as a natural sentence with a high degree of coincidence.
- the first parallel translation judging means excludes a predetermined frequently occurring word from the counting target power when counting the number of matching words, for example, as described in claim 6. This eliminates the effects of frequent words on the number of matching words, and determines the number of matching words to determine the degree of matching between the natural sentence and the original sentence. It can be used as an index that reflects more accurately.
- the first parallel translation judging means uses single or plural or tense in counting the number of matching words or mismatching words. It is preferable to count words whose endings are different due to the difference as matching words. For words whose endings are different due to single or tense differences, for example, the words are registered separately in a table, and words whose only endings do not match are registered in the table. It can be recognized by judging or not. As a result, it is possible to exclude the influence of words having different endings due to single or plural differences in tense, which should be regarded as matching words, on the number of matching words and the number of non-matching words.
- the number and the number of unmatched words can be used as an index that more accurately reflects the degree of matching between the natural sentence and the original sentence.
- differences between uppercase and lowercase letters of words in English sentences and the like are regarded as a match word.
- the first bilingual judging means duplicates a matching word appearing a plurality of times when counting the number of matching words as described in claim 8, for example. It is preferable not to count. As a result, it is possible to exclude the influence of the matching word appearing a plurality of times on the number of matching words, and to use the number of matching words as an index that more accurately reflects the degree of matching between the natural sentence and the original sentence. Further, instead of not performing the duplicate counting as described above, the matching word appearing multiple times may not be counted n or more (n ⁇ 2) predetermined times.
- the first bilingual judging means compares the natural sentence extracted by the search by the search means with the original sentence. It is preferable to evaluate the degree of similarity in the order of words and evaluate the degree of coincidence with the original text so that the higher the degree of similarity in the order of words, the higher the degree of coincidence with the original.
- the degree of similarity in the order of words may differ depending on the arrangement order.
- the similarity of the word arrangement order is evaluated, and the similarity of the word arrangement order is determined.
- the first bilingual judging means includes, for example, a natural sentence extracted by a search by the search means as described in claim 10. Then, the number of unmatched words existing between the matched words with the original sentence is counted, and the matching with the original sentence becomes higher as the number of unmatched words existing between the counted matched words becomes higher. It is preferable to evaluate the degree. As a result, the accuracy of the matching score is improved, and a natural sentence close to the original sentence can be selected with high accuracy based on the matching score.
- a recognizing means for recognizing a frequently appearing phrase which appears frequently in the same sentence in the source language as the translation target phrase is further provided, and the first bilingual judging means includes a natural language sentence extracted by the searching means.
- the specific frequent words recognized by the recognition means and present in the original sentence and the bilingual sentence of the natural sentence that contains the words to be translated the words to be translated and the specific frequent words are included. It is preferable to recognize a high-frequency bilingual translation of a phrase to be translated in a natural-language bilingual sentence, and determine the recognized high-frequency bilingual translation as a bilingual translation of the translated phrase in the original text.
- the frequency of occurrence (correlation) between a word to be translated and the same language in the source language is high. If a specific frequently-occurring word exists in the original text, the translation of the word to be translated is It is highly likely that the phrase and the specific frequent phrase are included, respectively, and correspond to the translation of the translation target phrase in the natural sentence. However, although it is highly likely that the appropriate translation of the translation target phrase in the natural language sentence that contains the translation target phrase and the specific frequently occurring phrase is the same, the above natural sentence stored in the storage means is high. It is possible that some of the natural sentences in which the appropriate translation of the target phrase is different.
- the invention according to claim 11 recognizes a frequent phrase having a high correlation with the phrase to be translated, and, among the recognized frequent phrases, a specific frequent phrase existing in the original text and the phrase to be translated are each
- a high-frequency bilingual translation is recognized for the translation target phrase in the natural sentence bilingual sentence that includes each of the translation target phrase and a specific frequently appearing phrase. Recognized high-frequency translations are interpreted as the translations of the words to be translated in the original text.
- the above frequently-recognized phrases can be recognized by, for example, registering the phrases frequently occurring in the same sentence of the source language in a table and referring to the table.
- the trouble of creating the table can be omitted and the table can be saved.
- An effect is also obtained that the storage capacity required for storing can be reduced.
- the target word phrase which is present in the original sentence and not included in the natural sentence extracted by the search by the search means can be substituted.
- Judgment means for judging an alternative word is further provided, and the first bilingual judging means includes, among natural sentences extracted by the search by the search means, the alternative word and the translation target word judged by the judgment means, respectively. It is preferable to determine at least the translation of the translation target phrase in the natural language bilingual translation as at least the translation target phrase in the original text.
- a natural sentence in which a specific word in the original sentence is replaced by another word may be extracted by the search by the search means.
- the meaning of the original sentence and the natural sentence If the meanings are similar, it is considered that one word and another word are interchangeable.
- a specific phrase can be substituted for the original sentence. If there is a natural sentence that has been replaced with another word (alternative phrase) that has a similar relationship, this natural sentence is likely to be similar in meaning to the original sentence, so this natural sentence is selected. It is desirable.
- an alternative word that can be substituted for the word of interest! / ⁇ that is present in the original sentence and included in the natural sentence extracted by the search by the search means is used. Judgment is made by the judgment means, and the first bilingual judging means is a bilingual sentence of the natural sentence extracted from the natural sentence extracted by the search means and containing the alternative word and the translation target word judged by the judgment means.
- the determination of the substitute words by the determining means according to the twelfth aspect of the present invention is performed, for example, by registering words having a replaceable relationship in a table and referring to this table.
- a natural sentence including the target word is searched from a plurality of natural sentences stored in the storage means, and the natural sentence extracted by the search is searched for.
- a natural sentence having the same syntax is stored in the storage means, and a search is performed from among the natural sentences, and the natural sentence extracted by the search is replaced with the word of interest! / Puru is determined as an alternative word. You may do so. Also in this case, the time and effort for creating the table can be omitted, and the storage capacity required for storing the table can be reduced.
- the bilingual determination device as set forth in claim 14 is characterized by that A storage unit for storing a plurality of sentences in association with a bilingual sentence of the target language; and a plurality of natural sentences of the source language stored in the storage unit, the words to be translated in the original sentence of the source language are included.
- a high-frequency bilingual translation of the translation-target phrase in the natural-language bilingual sentence including each of the translation target phrase and the specific frequent phrase is recognized, and the recognized high-frequency translation is translated.
- a second translation determining means for determining at least translation of the translated phrase in the serial textual is configured to include a.
- the same storage means and search means as those of the first aspect are provided, and the recognizing means performs translation based on a natural sentence extracted by the search by the search means. Recognize frequently occurring words and phrases that appear frequently in the same sentence of the target word and the source language, and the second bilingual judging means recognizes the natural sentence extracted by the search By translating a natural sentence that contains a phrase to be translated and a specific frequently appearing phrase by referring to a specific frequently occurring phrase and a bilingual sentence of a natural sentence that contains each of the words to be translated. A high-frequency bilingual translation is recognized for the target phrase, and the recognized high-frequency bilingual translation is determined as a bilingual translation of the target phrase in the original text.
- the phrase to be translated may include a plurality of words, but when the number of words constituting the phrase to be translated increases, the search means However, there is a possibility that a natural sentence that contains all the words to be translated is not extracted even if a search is performed. Considering this, for example, as described in claim 15, When the phrase to be translated is composed of a plurality of words, the search means includes at least one of the plurality of words constituting the phrase to be translated from among the plurality of natural sentences of the source language stored in the storage means.
- a natural sentence that has been included that is, search for a natural sentence that includes all the words to be translated and a part of the natural sentence and a part of the word to be translated!.
- search for a natural sentence that includes all the words to be translated and a part of the natural sentence and a part of the word to be translated! a phrase composed of many words is specified as the phrase to be translated, so that all the phrases to be translated are included! /
- the natural sentence is stored in the storage means!
- it is possible to obtain a natural sentence (translated sentence) from which at least a translation of the phrase to be translated can be estimated by the search means.
- the translation judging method according to the invention according to claim 16 is characterized in that a natural language sentence in a source language composed of a plurality of words, each of which is stored in a storage unit in association with a bilingual sentence in a target language.
- the translation judging method according to the invention according to claim 17 is characterized in that a natural language sentence in a source language composed of a plurality of words, each of which is stored in a storage unit in association with a translated sentence in a target language.
- the words to be translated in the translated words of the natural sentences and the specific frequently-used words are respectively included. It recognizes the high frequency translation of the attached, a translation of the recognized high frequency, comprising a third step of determining a translation of the translated phrase in the original sentence, Runode invention of claim 14, wherein
- an appropriate bilingual translation of the phrase to be translated in the original sentence (a high possibility that a natural translated sentence can be obtained from the original sentence as a sentence in the target language) can be obtained.
- a degree of matching between a sentence and the original sentence is obtained, and based on the obtained degree of matching, at least a translation of the translation target phrase in the selected translated sentence of the natural sentence is compared with a translation of at least the translation target phrase in the original sentence. It functions as the first means of judging translation.
- a program according to the invention as set forth in claim 18 is a computer connected to a storage means for storing a plurality of natural sentences of a source language consisting of a plurality of words in association with a bilingual sentence of a target language. And the other computer connected to the storage means via a communication line.
- the above-mentioned search means and the first computer may be used. Since the computer executes the program according to the invention described in claim 18, the computer functions as the translation determination device described in claim 1. In the same way as in the invention described in claim 1, it is possible to obtain an appropriate bilingual translation of the word to be translated in the original sentence (a high possibility that a natural translated sentence can be obtained as the target language sentence from the original sentence). Can.
- a recognition unit for recognizing a frequently appearing phrase that appears frequently in the same sentence of the source language and the translation target phrase, and a natural sentence extracted by the search by the search unit.
- the specific phrase frequently recognized in the original text and the phrase to be translated are each included by referring to the bilingual sentence of the natural sentence, and the translation target phrase and the specific frequent phrase are referred to.
- the translation target words in the bilingual sentences of the natural sentences included in each! Recognize all the high-frequency translations and make the recognized high-frequency translations function as second translation determination means for determining the translations of the words to be translated in the original text.
- a program according to the invention described in claim 19 is a computer connected to a storage means for storing a plurality of natural sentences of a source language composed of a plurality of words in association with a bilingual sentence of a target language. Or the other computer connected to the storage means via a communication line, and the above-mentioned search means and recognition means. Since the computer is a program for causing the computer to function as the second translation judging means, the computer functions as the bilingual judgment device according to claim 14 by executing the program according to the invention described in claim 19. Thus, similar to the invention described in claim 14, appropriate translation of the words to be translated in the original sentence (the possibility that a natural translated sentence can be obtained from the original sentence as a sentence in the target language, ) Can be obtained.
- the invention's effect is a computer connected to a storage means for storing a plurality of natural sentences of a source language composed of a plurality of words in association with a bilingual sentence of a target language. Or the other computer connected to the storage means via a communication line, and
- the present invention searches for a natural sentence including a phrase to be translated in a source language original sentence from a plurality of source language natural sentences stored in a storage unit,
- the degree of coincidence between the natural sentence extracted by the search and the original sentence is obtained, and at least the translation of the target phrase in the bilingual sentence of the natural sentence selected based on the obtained degree of coincidence is determined with the translation of at least the target phrase in the original sentence. Since the judgment is made, there is an excellent effect that a natural bilingual sentence is likely to be obtained as a target language sentence from the original sentence of the source language, and an appropriate bilingual translation of the word to be translated in the original sentence can be obtained.
- the present invention searches a plurality of natural sentences of the source language stored in the storage means for a natural sentence including the translation target phrase in the original sentence of the original language. Based on the extracted natural sentence, it recognizes frequently-used words and phrases that appear frequently in the same sentence in the source language and the target words to be translated, and includes the specific frequently-used words and target words in the original sentence.
- the high-frequency bilingual translation of the natural language bilingual sentence that includes the word to be translated and a specific frequently-used word is recognized and the recognized Since the translation of the frequency is determined as the translation of the phrase to be translated in the original sentence, there is a high possibility that a natural translated sentence can be obtained from the original sentence of the original language as a sentence of the target language.
- Brief description of the drawings which has an excellent effect that an appropriate translation of the words to be translated in the original text can be obtained.
- FIG. 1 is a block diagram showing a schematic configuration of a PC according to the present embodiment.
- FIG. 2 is a flowchart showing the contents of a translation determination process.
- FIG. 3 is a chart showing an example of a natural sentence and a bilingual sentence including a “ru operation” registered in the bilingual DB.
- FIG. 4 is a block diagram for explaining an embodiment in which the present invention is applied to a computer system in which a client PC and a server “computer are connected” via a network.
- FIG. 5 is a block diagram showing an example of a schematic configuration of a machine translation device to which the present invention has been applied.
- FIG. 1 shows a personal computer (PC) 10 that can function as the electronic dictionary device described above.
- the PC 10 includes a CPU 10A, a ROM 10B, a RAMIOC, and an input / output port 10D, which are connected to each other via a bus 10E such as a data bus, an address bus, or a control bus.
- the input / output port 10D has various input / output devices such as a CRT, LCD, and other displays 12, a keyboard 14 for the user to input data, a mouse 16, a hard disk drive (HDD) 18, and a CD-ROM M24.
- a CD-ROM drive 20 for reading object data and a scanner 22 for reading paper documents are connected to each other.
- the PC 10 has a translation determination program (corresponding to the program described in claims 18 and 19) for causing the PC 10 to function as an electronic dictionary device installed in the HDD 18.
- a bilingual database (bilingual DB) storing data used by the bilingual determination program for the bilingual determination is also stored.
- the bilingual DB is also stored in the HDD 18 by, for example, prerecording it in the CD-ROM 24 and configuring the setup program so that it is simultaneously written to the HDD 18 when the bilingual judgment program is installed. Can be done.
- the bilingual DB contains a natural sentence composed of multiple words and described in the source language (any of unedited sentences, phrases, collocations, fixed expressions, collocation, etc. Or, the text data of a specific example will be described later.
- the translated text is written in the target language (this translated text is also a natural sentence that has not been edited or processed such as word-by-word division or polysemy extraction).
- the HDD 18 that stores the bilingual DB corresponds to the storage unit according to the present invention.
- the bilingual DB can be recorded on a recording medium such as a CD-ROM 24 or a DVD-ROM, and can be used by directly reading out data from the recording medium.
- the recording medium on which the DB is recorded functions as the storage unit according to the present invention.
- the words in the source language and the translations in the target language are also registered in correspondence.
- the bilingual DB according to the present embodiment can be created, for example, by appropriately adding a natural sentence and its bilingual sentence to an existing dictionary in which words in the source language and bilinguals in the target language are associated. .
- a phrase (a phrase to be translated: also a word, or a phrase composed of a plurality of consecutive words in the source text) in which a translation in the target language is found in the source text described in the source language.
- the user recognizes that there is (1), the user performs a predetermined operation to output a bilingual translation of the translation target word in the original text via the PC 10.
- the original text is text read into the PC 10 as text data (for example, text entered by the user through the keyboard 14, text created by word processing software, text in a web page being browsed via the Internet, etc.).
- OCR Optical Character Recognition
- Characters obtained by optical recognition (character recognition) processing can be applied.
- the above-mentioned predetermined operation includes, for example, selecting the word to be translated while the original text is displayed on the display 12, highlighting the word to be translated, and then right-clicking the word to be translated. By doing so, it is possible to apply an operation such as selecting an item corresponding to “translation output” in the displayed context menu.
- the translation determination program shown in FIG. 2 is performed by executing the translation determination program by the CPU 10A of the PC 10.
- This translation determination process is a process to which the translation determination method described in Claims 16 and 17 is applied, and by performing this process, the PC 10 can use the electronic dictionary device (described in Claims 1 and 14). Function as a parallel translation judging device).
- step 100 the text data of a single original sentence (the original sentence to be processed) including the specified phrase to be translated is imported, and the translation in the imported original sentence to be processed is performed. Capture information that identifies the target phrase.
- the original sentence to be processed may be a sentence containing the phrase to be translated, or may be a phrase, collocation, fixed expression, or collocation containing the phrase to be translated. If any of the collocations is used as the original text to be processed, the user can specify the phrase, collocation, fixed-form expression, and collocation as the original text to be processed, or the translation processing automatically performs the translation. Judgment is also possible.
- step 102 using the text data of the original text to be processed fetched in step 100 as a key, a natural sentence registered in the bilingual DB that completely matches the original text to be processed is selected. While searching for sentences, the text data of the words to be translated is used as a key, and registered in the bilingual DB, the words to be translated are included in the natural sentences! /, The natural sentences (at least including the words to be translated) , Search for a part of the original sentence! Also, if the target phrase is composed of multiple words, Natural sentences containing at least one of the words that make up the phrase are also searched.
- Step 102 corresponds to the retrieval means described in claims 1 (specifically, claims 2 and 15), claims 14, 18, and 19, and the first step described in claims 16 and 17 Steps are also supported.
- step 104 and subsequent steps correspond to the first translation judging means described in claims 1 and 18.
- step 104 it is determined whether or not a natural sentence that completely matches the original sentence to be processed by the search in step 102 is extracted from the bilingual DB. If this determination is affirmative, the process proceeds to step 106, where the bilingual sentence of the target language registered in the bilingual DB is read out in association with the natural sentence that completely matches the original sentence to be processed, and displayed on the display 12. Is displayed, and the process ends.
- the bilingual translation of the phrase to be translated on the read bilingual sentence is recognized, and the bilingual translation of the recognized phrase to be translated is highlighted.
- the user can recognize an appropriate bilingual translation of the specified phrase to be translated (highly likely to obtain a natural bilingual sentence from the original sentence as a sentence in the target language, bilingual translation). It is possible to recognize an appropriate bilingual sentence of the original text (a natural bilingual sentence as a sentence in the target language).
- the above steps 104 and 106 correspond to the first translation judging means according to claim 2.
- Example 1 The above processing will be further described by way of examples.
- the source language is English and the target language is Japanese.
- the source language and the bilingual translation are registered in units of words, so the word “safety” that constitutes the above translation target phrase is The translation “safe” is selected for ⁇ .
- the natural sentence of the source language and the bilingual sentence of the target language are registered in the bilingual DB in association with each other. May have been registered in the translation DB. If a natural sentence that matches exactly is registered in the bilingual DB, the bilingual sentence registered in the bilingual DB in association with the natural sentence is highlighted as follows: Is output.
- a natural bilingual sentence can be obtained as a native language.
- a natural sentence that completely matches the original sentence to be processed is added to the sentence, and partially matches the original sentence to be processed extracted by the search in step 102! / You may also display natural texts together!
- a plurality of natural sentences that partially match the original sentence to be processed are extracted by search, but when displaying them, the following describes each natural sentence that partially matches the original sentence to be processed. It is preferable to calculate the degree of coincidence and display the bilingual sentences of each natural sentence on the display 12 in descending order of the degree of coincidence.
- step 104 determines whether the natural sentence that completely matches the original sentence to be processed is not extracted by the search in step 102. If the natural sentence that completely matches the original sentence to be processed is not extracted by the search in step 102, the judgment in step 104 is denied and the process proceeds to step 108.
- the natural sentence extracted from the bilingual DB by the search of 102 (partially matched with the original sentence to be processed! /, Each natural sentence), one of the natural sentence words in the original sentence The number of words that match (the number of matching words) is counted, and the degree of matching with the original text is calculated based on the counting result of the number of matching words.
- V in the word for example natural sentence of English to be frequent in the natural sentence of the original language, "th e", "t 0", "in” , etc.
- the counting of the number of matching words in step 108 is performed with reference to the frequent word table, and the words registered in the frequent word table are excluded from the counting target words of the number of matching words. As a result, it is possible to eliminate the effect of frequently appearing words on the number of matching words.
- the above processing in step 108 corresponds to the first translation judging means according to claim 6! / ,.
- the HDD 18 when the bilingual translation determination program is installed, the HDD 18 also stores an inflection-changed word table in which words whose inflections differ due to a difference in tense or tense are registered. Then, in the counting of the number of matching words in step 108, if a word whose only the ending does not match appears, the ending change word table is referred to determine whether the ending mismatch is due to a single or multiple tense difference. Is judged and the difference of single or tense Thus, words having different endings are counted as matching words.
- step 108 corresponds to the first translation judging means according to claim 7.
- step 108 when counting the number of matching words in step 108, even if a word once included in the number of matching words due to matching with any word in the original sentence appears again in the natural sentence, By not counting the number of matched words, duplicate words that appear multiple times are not counted. As a result, even when the same matching word exists in a plurality of places in a natural sentence, it is possible to eliminate the influence of the matching word on the number of matching words.
- the above processing in step 108 corresponds to the first translation judging means described in claim 8.
- the arithmetic expression of the matching degree can be determined so that the matching degree increases as the number of matching words increases.
- the number of matching words is normalized by the number of words constituting the translation target phrase.
- the above step 108 corresponds to the first translation judging means described in claim 3 (specifically, claim 4).
- step 110 by comparing the degrees of matching calculated for each natural expression, it is determined whether or not there is a plurality of natural sentences having the highest degree of matching, and whether or not there is power. If the determination is negative, the process proceeds to step 112, where the bilingual sentence of the target language registered in the bilingual DB associated with the natural sentence having the highest matching degree is read, and the translation target on the read bilingual sentence is read.
- the bilingual translation of the word is recognized, and the read bilingual sentence is displayed on the display 12 so that the bilingual translation of the recognized translation target phrase is highlighted, and the process ends.
- the user can recognize an appropriate bilingual translation of the specified phrase to be translated (the possibility of obtaining a natural bilingual sentence from the original text as a sentence in the target language is high).
- step 112 among the natural sentences extracted by the search in step 102, in addition to the natural sentence with the highest matching degree, a plurality of natural sentences (in the descending order of the matching degree, A bilingual sentence of a number of natural sentences or all natural sentences having a matching degree equal to or more than a predetermined value is also read out from the bilingual DB, and is displayed on the display 12 as a list. If multiple natural sentences with the same degree of coincidence exist, the number of words that do not match the original sentence in each natural sentence is counted, and when displaying a list of translated sentences, the natural sentence with the same degree of coincidence is displayed.
- the bilingual sentences are displayed in ascending order of the number of unmatched words in the corresponding natural sentence.
- Example 2 The above process will be further described by way of an actual example.
- the source language is English and the target language is Japanese.
- the natural sentence (1)-(8) shown in Fig. 3 is a translated sentence (1) If it is registered in the bilingual DB in association with (8), except for the natural sentence (3) shown in Fig. 3, the number of words that match the original sentence is “1”, the degree of matching is 100%, The sentence (2) “operations” is also counted as a matching word as described above), and the natural sentence (3) has the number of matching words with the original sentence is “4” (the word ⁇ the ⁇ is a frequent occurrence and the number of matching words is counted) (Excluded from the target) and the matching score becomes ⁇ 00%, so it is associated with the natural sentence (3) and registered in the bilingual DB, and the translated sentence (3) is translated as follows: The translation of is highlighted and output.
- Example 3 The source language is English and the target language is Japanese.
- ⁇ ⁇ operation ⁇ included in the original sentence to be processed is specified, and the natural sentence (1)-(8) shown in Fig. 3 is a translated sentence (1) (1)
- the number of words that match the original sentence S (l) is 100%, and the natural sentence (2) matches the original sentence
- the matching score becomes ⁇ 00%, so it is correlated with the natural sentence (2) and the translation DB
- the bilingual sentence (2) is output with the bilingual translation of the translation target phrase highlighted as follows.
- Example 4 It is difficult to determine the unit for which a Chinese sentence is required to be translated when translating into a sentence in another language that is difficult to determine if the Chinese sentence is not familiar with Chinese. There are many things.
- natural sentences that include the specified translation target phrase are searched for from among the natural sentences registered in the bilingual DB (when there are multiple translation target phrases).
- the natural sentence that contains at least one of the words that make up the phrase to be translated is also searched at the same time.
- the sentence is displayed (as well as the bilingual sentence of the natural sentence other than the natural sentence with the highest matching score). Recognize appropriate translations at the same time.
- the words to be translated are “Departure”, “Development”, “Developing”, Regardless of which of "Developing China” and "Developing National” is specified, bilingual translations of the following natural sentences are listed in the following order based on the degree of coincidence.
- the degree of matching when "nation” is specified is also shown, but it goes without saying that the degree of matching varies depending on the word specified as the word to be translated.
- each natural sentence has a matching word number of "1" and a matching power of 100%.
- the original text is It is displayed in the following order, which is different from the case of "nation".
- step 110 If a plurality of natural sentences having the highest degree of matching exist! /, The determination in step 110 is affirmed, and the process proceeds to step 114.
- steps 114 and 116 the translation target word and the original Natural sentences are selected based on words that appear frequently in the same sentence of a word (words with high correlation with the words to be translated: equivalent to frequent words described in claims 11, 14, 17, and 19) Perform correlation analysis.
- step 114 for each natural sentence whose bilingual DB power is also extracted by the search in step 102, each word other than the phrase to be translated in the original sentence to be processed (however, registered in the frequent word table, By retrieving whether or not the word to be translated is included in the same sentence in the source language and the language to be translated, it is possible to search for words with high frequency (high correlation with the phrase to be translated, (Word) is included in the original sentence to be processed. In this process, for example, among words other than the words to be translated in the original text to be processed, for example, Words that appear in each natural sentence more than a predetermined number of times (either once or multiple times! ⁇ ) can be determined to be words with high correlation with the words to be translated.
- step 114 corresponds to the recognizing means described in claims 11 and 14, and also corresponds to the second step described in claim 17.
- step 116 the height of the correlation with the phrase to be translated and the presence or absence of the word are determined by the search in step 114. If the determination is affirmative, the process proceeds to step 118, and among the natural sentences for which the bilingual DB power was also extracted by the search in step 102, the correlation between the translation target phrase and the translation target phrase determined in step 114 is high. , A word (existing in the original sentence to be processed! / ⁇ word) is included! / ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ! Judge high frequency translations of the target phrase.
- a high-frequency bilingual translation of a phrase to be translated specifically has a high correlation with the phrase to be translated and the phrase to be translated!
- the translation of the translation target phrase in the natural sentence is determined to be a high-frequency translation. If a plurality of natural sentences are extracted as the above natural sentences, the translation in these natural sentences is performed. Among the bilingual translations of the target phrase, the bilingual translation with the highest frequency of occurrence can be determined as the high-frequency bilingual translation.
- the natural sentence that includes the word to be translated, has a high correlation with the word to be translated, and includes the same word as the word to be processed as the word to be processed.
- the natural sentence is likely to be a sentence that uses the word to be translated with the same meaning as the original sentence to be processed, but in the above natural sentence, the translation for the word to be translated is different. There may be mixed natural sentences.
- the translation of the natural sentence is referred to by referring to the bilingual sentence of the natural sentence which has a high correlation with the word to be translated and the word to be translated. Since a high-frequency translation of the target phrase is determined, it is possible to obtain an appropriate bilingual translation of the target phrase in the original text to be processed.
- step 119 the translation target phrase is associated with the high-frequency bilingual translation recognized in step 118 in the natural sentence including the translation target phrase and the word having a high correlation with the translation target phrase.
- Is registered in the bilingual DB Reads the bilingual sentence in the target language and displays the read bilingual sentence on the display 12 so that the translation target word on the read bilingual sentence and the bilingual translation with high correlation with the translation target phrase are highlighted. And terminate the processing.
- the user can recognize an appropriate bilingual translation of the specified translation target phrase (a bilingual translation that is likely to yield a natural bilingual sentence from the original sentence as the target language sentence).
- Step 116 to step 119 described above are the first translation judging means described in claim 11, the second translation judgment means described in claims 14 and 19, and the third translation step described in claim 17. Respectively.
- step 119 of the natural sentences extracted by the search in step 102, the words to be translated and words having a high correlation with the words to be translated are included, and the words to be translated are recognized.
- multiple natural sentences in descending order of matching a certain number of natural sentences in descending order of matching, or all natural sentences with a matching score equal to or greater than a predetermined value
- the bilingual sentences are also read from the bilingual DB, and are listed on the display 12 in the order according to the degree of coincidence and the number of unmatched words.
- Example 5 The above process will be further described by way of an actual example.
- the source language is English and the target language is Japanese.
- ⁇ ⁇ operation ⁇ ⁇ included in the original sentence to be processed is specified, and the natural sentence (1)-(8) and other natural sentences shown in FIG. If each of the natural sentences (6) and 8) is registered in the bilingual DB with the bilingual sentence, the number of matching words in the original sentence is "2". It is difficult to choose).
- a natural sentence containing the word “operation” to be translated using each word eg, “remove”, “rectal”, “cancer”, etc.
- the natural sentence (6) (8) contains "cancer ⁇ ", which means that words with a high correlation with the target phrase ⁇ operation "in the original sentence to be processed Is extracted as "cancer".
- a natural sentence containing the word to be translated “operation” and the word “cancer” having a high correlation is used by using the word to be translated “operation” in the same meaning as the original sentence to be processed.
- Natural sentence that is likely to be a sentence that meets the above conditions registered in the bilingual DB There is a possibility that some of the words to be translated are different from the original sentence to be processed!
- the natural sentence (8) corresponds to it.
- the natural sentence (8) contains the word "operation” to be translated and the correlation is high and contains the word "cancer”
- the translation of the word ⁇ operation ⁇ to be translated is "operation", so the original text to be processed It is different.
- the bilingual sentence of the natural sentence is referenced.
- the translation target phrase “operation” and the correlation are high, the word “cancer” is included, and the translation target phrase “operation” is associated with the translation “operation”.
- the sentence is registered in the bilingual DB, the word "operation” and the word “cancer” are included in the word "cancer”!
- the high-frequency bilingual translation is judged to be “surgery,” and the translation target word ⁇ operation ”, the correlation is high !, and the word“ cancer ”are included.
- the corresponding natural sentence (6) is selected and associated with the natural sentence (6) and registered in the bilingual DB, and the translated bilingual sentence (6) is translated as shown below.
- the bilingual word with the highest word is highlighted and output.
- an appropriate natural sentence (a bilingual sentence) is selected using the word to be translated ⁇ operation ⁇ in the same meaning as the original sentence.
- step 116 determines whether a word having a high correlation with the phrase to be translated is not extracted. If a word having a high correlation with the phrase to be translated is not extracted, the determination in step 116 is denied and the process proceeds to step 120. In steps 120 to 134, the original text to be processed is processed. Based on the words of interest other than the words to be translated and alternative words that can be substituted (corresponding to the alternative words described in claim 12), a scheme analysis for selecting a natural sentence is performed.
- step 120 a word (referred to as a word of interest) that is present in the original sentence to be processed and is not present in each natural sentence extracted from the bilingual DB by the search in step 102 is determined.
- step 122 a natural sentence containing the word of interest is searched from the natural sentences registered in the bilingual DB. Note that there may be a plurality of target words. In this case, the search in step 122 is performed for each target word.
- step 124 only the noticed word in each natural sentence extracted by the search in step 122 is different, and the natural sentence (the natural sentence extracted by the search for the natural sentence including the noticed phrase described in claim 13)
- the search condition is set to search for each natural sentence with the same syntax as the sentence), and based on the set search conditions, the corresponding natural sentence is searched from among the natural sentences registered in the bilingual DB. .
- next step 126 it is determined whether or not the corresponding natural sentence has been extracted by the search in step 124. If the same natural sentence exists in the bilingual DB except that the target word is replaced with another word, and the same natural sentence exists in the bilingual DB, the another word may be a substitute word that can be used in place of the target word. Is high. For this reason, when the determination is affirmed, the process proceeds to step 128, and the natural word extracted by the search recognizes the word replacing the target word as an alternative word of the target word. Steps 120-130 described above correspond to the determination means described in claim 12 (specifically, claim 13).
- step 130 there is a natural sentence in which the word of interest is replaced with an alternative word among a plurality of natural sentences having the maximum matching score (instead of this, the matching score is equal to or more than the predetermined value). Is determined. If there is a natural sentence that includes an alternative word in place of the target word in the original sentence to be processed among the multiple natural sentences with the highest matching score! / ⁇ , the natural sentence Is likely to be a sentence that uses the word to be translated with the same meaning as the original sentence to be processed. Therefore, if the determination is affirmative, the process proceeds to step 132, The maximum degree (instead of this, it may be “more than the predetermined value of the degree of coincidence”!
- step 132 among the natural sentences extracted by the search in step 102, in addition to the natural sentences that have the highest matching degree and include an alternative word instead of the word of interest, descending order of matching degree
- bilingual sentences of a plurality of natural sentences (a certain number of natural sentences in descending order of coincidence, or all natural sentences whose coincidence is equal to or more than a predetermined value) are also read from the bilingual DB, and are read in the order according to the degree of coincidence and the number of mismatched words A list is displayed along with the display 12.
- Example 6 The above process will be further described by way of an actual example.
- the source language is English
- the target language is Japanese
- ⁇ have ⁇ is specified as the target phrase for translation in the source text including “have lunch ⁇ ”.
- the target phrase "natural sentence using the word” have "in the same meaning as the original sentence to be processed) is registered, the natural sentence containing" have lunch "is not registered, and the judgment based on the degree of coincidence ⁇ correlation If the analysis fails to identify an appropriate bilingual sentence, the words to be translated ("have") in the source text to be processed and words other than the words judged to have high correlation with the words to be translated are high.
- Each word (e.g., "lunch”) is taken as a word of interest, and a natural sentence containing each word of interest is searched, whereby, for example, for the word of interest "lunch ⁇ ", for example, "eat lunch” or "take” Natural sentences including "a late lunch” are extracted.
- a natural sentence in which only the focused word in each natural sentence is different is searched.
- a natural sentence extracted as a natural sentence including the word of interest "lunch”! Is a natural sentence in which only the word of interest is different, such as "eat breakfast ⁇ " or "take a late breakfast ⁇ ".
- the word of interest ⁇ lunch ⁇ ⁇ is replaced with the alternative word ⁇ breakfast ⁇ , and is associated with the natural sentence (natural sentence including “have breakfast ⁇ ) and registered in the bilingual DB.
- the translation of the word to be translated ⁇ have ⁇ ("eat") is highlighted, and the translation of the alternative word "breakfast"("breakfast") is marked and output. Therefore, in this example, an appropriate natural sentence (a bilingual sentence) is selected by using the word to be translated ⁇ have ⁇ in the same meaning as the original sentence to be processed.
- the translated text registered in the DB is highlighted, and the translated text (“Kai”) of the target phrase ⁇ have ⁇ is highlighted.
- a translation (“dog") of the alternative word "dogs” is marked and output. Therefore, in this example, too, an appropriate natural sentence (a bilingual sentence) is selected by using the word to be translated "have ⁇ " with the same meaning as the original sentence to be processed.
- Example 7 The source language is Japanese, the target language is English, and “Kake” is specified as the translation target phrase in the original sentence “Hot hot water” to be processed. ”(Using the word“ kake ”for translation with the same meaning as the original sentence to be processed), but“ kake hot water ”is not registered. Since other natural sentences are also registered, such as ⁇ Make a difference '', judgment based on the degree of coincidence ⁇ If it is not possible to identify an appropriate bilingual sentence even by correlation analysis, it will be in the original sentence to be processed "Hot water” is set as the word of interest, and a natural sentence containing the word of interest is searched for. This allows, for example, “soak in hot water”, “pour hot water”, “pour hot water”, “wash with hot water”, and “draw from hot water”. Natural sentences such as "Kikari” and "Hot with hot water” are extracted.
- the word “water” has a large number of appearances, and there is also a natural sentence “sprinkle water”. Therefore, the word “water” replaces the target word “hot water” in the original text “hot water”. It can be determined that the alternative word is likely to be possible. For this reason, of the natural sentences extracted in the first search (search for natural sentences including the translation target phrase ⁇ ⁇ ⁇ ), the word of interest “hot water” is replaced with the alternative word “water”, and the natural sentence “water”
- the translations registered in the translation DB that are associated with "Kakeru” are displayed as follows, with the translation ("pour") of the translation target phrase "Kake” highlighted and the alternative word "Water” The translation ("water”) is marked and output.
- a word such as "3" is considered to be the word of interest, and Is likely to be a substitute for the word of interest "3", and is judged as an alternative word, so that the first search (search for natural sentences containing the target phrase "kake")
- search for natural sentences containing the target phrase "kake” In the natural sentence extracted in step 2, the target word "3” is replaced with the alternative word "4", and the natural sentence "4 multiply” is registered in the bilingual DB,
- the translation of the target phrase "Kake”("multiply") is highlighted, and the translation of the alternative word "4"("four") is marked and output.
- an appropriate natural sentence (a bilingual sentence) is selected using the word to be translated “kake” with the same meaning as the original sentence to be processed.
- step 134 all natural sentences extracted by the search in step 102, or!, Are a plurality of natural sentences in descending order of coincidence (a certain number of natural sentences, All natural sentences having a matching degree equal to or more than a predetermined value) are read in the bilingual sentences registered in the bilingual DB, and the bilingual translation of the translation target phrase on each read bilingual sentence is recognized and recognized.
- the translated sentence of people exit the identity and the list display and the processing in descending order of the matching degree up each natural sentence matching degree in order according to the number of mismatches word on the display 12.
- the user can recognize some suitable translation candidates for the specified phrase to be translated.
- the degree of coincidence is calculated using only the number of matching words
- the present invention is not limited to this.
- the degree of coincidence increases. Is determined and the degree of match is determined so that the degree of match increases as the number of mismatched words between the natural sentence and the original sentence decreases, and the degree of match is calculated and evaluated according to the number of matched words and the number of mismatched words. You can do it.
- This aspect corresponds to the invention described in claim 5. Initially, the degree of matching is evaluated based only on the number of matching words, and the correlation is evaluated. If it is difficult to select a single natural sentence (parallel translation) even by performing a parsing analysis or scheme analysis, the number of unmatched words is counted and the natural sentence (the bilingual sentence of ) May be selected.
- the similarity of the order of words in the natural sentence and the original sentence, or the natural sentence existing between matching words in the original sentence is also used to evaluate the matching so that the similarity of the order of words in the natural sentence and the original sentence increases as the similarity in the original sentence increases.
- the matching score may be evaluated so that the matching score with the original sentence increases as the number of unmatched words decreases.
- the aspect using the similarity of the order of words in the natural sentence and the original sentence corresponds to the invention described in claim 9, and the use of the number of non-matching words existing between matching words is described in claim 10
- the evaluation of the matching degree in consideration of the similarity of the order of the words in the natural sentence and the original sentence and the number of non-matching words in the natural sentence existing between the matching words in the original sentence is performed by, for example, the following processing. It can be realized by.
- a first evaluation value is assigned to each word in the original sentence according to the distance (the number of words) from the word to be translated in the original sentence.
- the first evaluation value for the word to be translated in the original sentence becomes the maximum, and the first evaluation value for each word other than the word to be translated in the original sentence is The value can be determined so that the value decreases as the distance from the phrase increases (as the number of intervening words increases).
- the fourth word D in the original sentence is the word to be translated
- the following first evaluation value can be assigned to each of the words A to J in the original sentence.
- the translation DB power is also extracted by searching in step 102 For each natural sentence that contains the target phrase to be translated, a matching word that matches any one of the words in the original sentence among the words in the natural sentence is recognized, and the first evaluation given to each recognized matching word first Calculate the sum of the values. Then, the total value of the first evaluation values is used as the degree of coincidence, and the bilingual sentence of each natural sentence extracted by the search is output in descending order of the degree of coincidence (the total value of the first evaluation values) of each corresponding natural sentence. . For example, if a natural sentence 1-a natural sentence 5 of the following word sequence is extracted by a search (however, The word X is any word),
- Natural sentence 1 (D, X, X, E, F, G)
- Natural sentence 4 (A, B, X, C, X, D)
- the first evaluation value is determined so that the value of each word other than the word to be translated in the original sentence increases as the distance from the word to be translated in the original sentence becomes smaller.
- each natural sentence containing the target phrase is evaluated based on the sum of the first evaluation values, so that more words exist in the original sentence near the target phrase.
- natural sentences that is, phrases (concatenated words) composed of words to be translated in the original sentence and words in the vicinity thereof, are likely to have high probability, and natural sentences (the similarity of the order of words in the original sentence is estimated to be high). Natural sentence) can be evaluated as a natural sentence with a higher degree of coincidence.
- the degree of coincidence (sum of the first evaluation values) of natural sentence 1 to natural sentence 4 is the same, and thus the degree of coincidence based on the first evaluation value is the same.
- the difference between the matching words that match any of the words in the original sentence and the words to be translated in each natural sentence (number of words)
- the second evaluation value is assigned to each natural sentence having the same degree of matching based on the first evaluation value.
- the second evaluation value for the translation target phrase included in each natural sentence is the largest, and the second evaluation value of the matching word other than the translation target phrase in each natural sentence is The value can be determined so that the value becomes smaller as the distance from the word to be translated becomes larger (as the number of intervening words increases).
- the bilingual sentences are output in descending order of the total value of the second evaluation values of the corresponding natural sentences (also included in the matching degree according to the present invention).
- the following second evaluation value is assigned to each matching word, and the total value of the following second evaluation values is obtained.
- the second evaluation value of the word to be translated is set to 10.0, and for other matching words, the number of words existing between the word to be translated is 0,1,2,3,4
- An example is shown in which the second evaluation value is set such that the second evaluation value decreases to 5.0, 2.0, 1.0, 0.5, 0.2, etc. as the calorie increases.
- the natural sentence 1-the natural sentence 4 (a bilingual sentence) is output in descending order of the total value (matching degree) of the second evaluation values, that is, in the order of the natural sentences 3, 2, 4, 1.
- the second evaluation value is determined as the distance between a matching word that matches any one of the words in the original sentence among the words in each natural sentence and the translation target phrase in each natural sentence becomes smaller.
- the natural sentence containing more words that match the original sentence Natural sentences with as few unmatched words as possible between words, that is, phrases (consequences) consisting of words to be translated in the original sentence and words around them! A natural sentence can be evaluated as having a higher degree of coincidence.
- the similarity evaluation in consideration of the similarity of the order of words in the natural sentence and the original sentence and the number of non-matching words in the natural sentence existing between matching words in the original sentence are performed, for example, by the following processing. It is also possible to realize this.
- a search is made to determine whether the previously extracted front word exists within a predetermined number of words (for example, within 3 words) from the reference position to the front side (for example, up to 3 words).
- a search is performed to determine whether or not the rear word has a reference position force also within the predetermined number of words (for example, within 3 words). Then, for the natural sentence in which the preceding word and the following word were found by the above search, a third evaluation value determined such that the value increases as the distance between the reference position and the preceding word and the following word decreases becomes smaller.
- a front word that has not been extracted and has a minimum distance from the translation target phrase in this case, a single word (between the previous target word and the translation target phrase)
- the previous word in which the original word exists is extracted, and the latter word that has not been extracted and has the minimum distance from the target phrase (in this case, the target word) is extracted from the latter word group in the original sentence.
- the previously extracted preceding word is within a predetermined number of words forward from the position of the preceding word found in the previous search.
- a third evaluation value is set so that the value increases as the distance from the position decreases.
- the position of the front word or the back word found in the previous search is the same as the position of the front word or the back word found this time.
- An evaluation value that is uniquely determined according to only the gap in the natural sentence may be used.However, in consideration of the gap between the word to be translated in the original sentence and the front or rear word to be searched, The effect of the side word or the back word on the third evaluation value increases as the distance between the word and the word to be translated in the original sentence becomes smaller.
- the third evaluation value may be determined such that the third evaluation value to be given becomes smaller as the distance from the word to be translated becomes larger.
- the value increases as the difference in the natural sentence between the position of the front word or the back word found in the previous search and the position of the front word or the back word found this time decreases.
- the third evaluation value increases as the distance between the above-described reference position and the position of the preceding word or the succeeding word found in the natural sentence decreases. You may set a third evaluation value to make it easier.
- word B is extracted as the front word
- word D is extracted as the back word.
- the original word B is extracted for each natural sentence including the translation target phrase extracted from the bilingual DB by the search. From the word within the predetermined number of words (for example, within 3 words), and whether the back word D is within the range of the specified number of words from the translation target phrase to the back (for example, within 3 words). Existence Therefore, a third evaluation value is assigned to the natural sentence in which the front word B and the back word D are found, respectively, in search of the power.
- word A is extracted from the original sentence as the front word and word E is extracted from the original sentence as the back word, and the preceding word A is extracted from the natural sentence in which the previous word B and the back word D were found in the previous search.
- word B is present within a predetermined number of words from the front word B to the front (for example, within three words)
- the rear word E is within the predetermined number of words to the rear word D (For example, 3 words or less) are searched for each other, and a third evaluation value is assigned to the natural sentence in which the front word A and the back word E are found.
- the order is as follows.
- the word X means an arbitrary word
- "Z" means a delimiter.
- the similarity evaluation in consideration of the similarity of the order of words in the natural sentence and the original sentence and the number of non-matching words in the natural sentence existing between the matched words in the original sentence are performed, for example, in the following processing. Therefore, it can also be realized.
- the front word having the smallest distance from the translation target phrase from the front word group existing before the translation target phrase in the original sentence (in this case, the front word adjacent to the translation target phrase) ) Is extracted, and for each natural sentence including the translation target phrase extracted from the bilingual DB by the search, the position where the translation target phrase exists in each natural sentence is defined as the reference position (in the natural sentence).
- the position where any of the words to be translated exists is set as the reference position), and the front word extracted earlier moves from the reference position to the front.
- a search is performed to determine whether or not a force exists within a predetermined number of words (for example, within three words).
- This process counts the number of words by distance (specifically, the first number of words by distance and the second number of words by distance) as the fourth evaluation value for each natural sentence containing the phrase to be translated.
- the distance between the reference position and the preceding word specifically, the number of unmatched words existing between the reference position and the preceding word
- was counted was calculated.
- the corresponding first number of words by distance is counted up.
- a front word that has not been extracted and has a minimum distance from the translation target phrase in this case, a single word (in the previous search.
- the previous word in which the previously extracted front word was found in the previous search was extracted. From the position in front of the word within a predetermined number of words (for example, within 3 words), and search for the natural sentence in which the previous word was found by the previous search. The distance between the position of the previous word and the previous word found this time is counted, and the distance between the previous word found in the previous search and the reference position is added to the counted distance, so that it is found by this search.
- the distance between the previous word and the reference position is calculated. The first distance by the number of words you each counted up. This process is repeated while extracting the front words from the original text in ascending order of the distance to the word to be translated, until there are no more front words that can be extracted from the original text.
- word C is extracted as the front word, and each word including the translation target phrase extracted from the bilingual DB is searched.
- Each natural sentence is searched for whether the preceding word B exists within a predetermined number of words (for example, within 3 words) from the target phrase in the natural sentence, and the preceding word C is found.
- the first number of words by distance is counted up. Table 1 below shows an example of the result of counting the number of words by distance at this point.
- word B is extracted as the front word in the original sentence, and the natural sentence in which the front word C was found in the previous search is within a predetermined number of words from front word C to the front (for example, 3 words).
- the search is performed to determine whether or not the front word B exists within (within) the first sentence, and the natural sentence in which the front word B is found is counted up by the first distance-based word count.
- Table 2 An example of the result of counting the number of words at the first distance at this point is shown in Table 2 below.
- word A is extracted as the front word, and the front word B is generated in the previous search.
- the presence or absence of the power of the front word A within the predetermined number of words (for example, within three words) from the front word B to the front was searched, and the front word A was found.
- a first count of the number of words for each distance is performed on a natural sentence. Table 3 below shows an example of the result of counting the first number of words by distance at this point.
- the distance between words is counted by a method different from the above first number of words by distance, and the second distance Separate Count as the number of words.
- the preceding word is not present in the original sentence adjacent to the target phrase in the original sentence group (the word C included in each natural sentence is not In this state, each natural sentence containing the translation target phrase extracted from the bilingual DB by extracting the front word with the minimum gap from the translation target phrase from the front word group in the original sentence in this state For each natural sentence!
- the position where the word to be translated exists is set as a reference position, and a search is performed to determine whether or not the extracted front word exists within a predetermined number of words (for example, within 3 words) from the reference position to the front. . Then, for the natural sentence in which the preceding word was found by the above search, the distance between the reference position and the preceding word (specifically, the number of mismatched words existing between the reference position and the preceding word) was counted and counted. The distance obtained by adding “1” to the distance is defined as the distance from the reference position, and the number of words according to the second distance corresponding to the distance from the reference position is counted up.
- a front word that has not been extracted and has the minimum distance from the phrase to be translated is extracted, and for each natural sentence in which the front word was found by the previous search, A search is performed to determine whether the extracted front word exists within a predetermined number of words (for example, within 3 words) from the position of the front word found in the previous search to the front, and the current search determines For the natural sentence where the word was found, the distance between the position of the front word found in the previous search and the previous word found this time is counted, and the counted distance is compared with the number of the previous word found in the previous search.
- a predetermined number of words for example, within 3 words
- the distance between the preceding word found in this search and the reference position is obtained, and the second number of words for each distance corresponding to the obtained distance is counted up. This process is repeated while extracting front words from the original text in ascending order of the distance from the phrase to be translated, until there are no more front words for which original text power can be extracted.
- word A—word E (A, B, C, D, E)
- the fourth word D in the original sentence is specified as the phrase to be translated.
- the front word C adjacent to the phrase to be translated in the front word group in the original sentence does not exist in the original sentence (the word C included in each natural sentence With the original sentence words B and A extracted in order and the distance from the reference position on each natural sentence is counted, as shown in Table 4 below, for example, The result of counting is obtained.
- each natural sentence is calculated. For each time, the counting results of the first number of words by distance and the second number of words by distance are compared, and if the inter-word distance is shorter among the first number of words by distance and the second number of words by distance, The counting result showing the result of (1) is selected as the final evaluation for the front word group.
- the natural sentence (A, B, C, D) has a distance of 0 and a number of words of 3 in the first number of words by distance shown in Table 3, whereas the first sentence shown in Table 5 In the number of words at distance 2 of 2, the number of words at distance 0 is 0, the number of words at distance 1 is 0, and the number of words at distance 2 is 2.
- A, B, C, D) is selected as the final evaluation for the front word group.
- the number of words at distance 0 in the first number of words by distance and the second number of words by distance differ. However, if the number of words at distance 0 is the same, the number of words at distances 1, 2, ... are compared sequentially, and the same distance between the first number of words by distance and the second number of words by distance. The one with more words is selected as the final evaluation.
- the second counting of the number of words according to the distance is based on the similarity of the order of words in the original sentence even in a natural sentence in which the order of some words in the original sentence is changed. This is to properly evaluate.
- natural sentence (C, A, B, D)
- the first number of words per distance is 0 words at distance 0, 0 words at distance 1, and 1 word at distance 2
- the second number of words by distance the number of words is 0 at distance 0 at distance
- the number of words is 2 at distance 1
- the number of words is 0 at distance 2.
- the second number of words at distance is the final evaluation of the preceding word group of the above natural sentence. Selected.
- the result of counting the number of words by the first distance is evaluated assuming that natural sentences (C, A, B, D) are (C, X, X, D), whereas The results of counting the number of words by distance are evaluated by considering the natural sentence (C, A, B, D) as ( ⁇ , ⁇ , ⁇ word C missing), D). Can be evaluated as being included in a more aggressive state.
- the final evaluation of the front word group is determined. Similar processing using the posterior word group that exists behind the target phrase (counting the number of first and second distance-based words, determining the final evaluation of the posterior word group) I do. Next, for each natural sentence, the final evaluation for the front word group and the final evaluation for the rear word group obtained are added and counted (the number of words by distance in each final evaluation is added for each same distance). Calculate the overall evaluation of each natural sentence. Then, based on the comprehensive evaluation of each natural sentence, the bilingual sentence of each natural sentence is output in the order of the bilingual sentence that indicates that the overall evaluation of the corresponding natural sentence indicates that the distance between words is shorter.
- the second counting of the number of words according to distance is not essential.
- the counting of the number of words by the second distance may be omitted, and the number of words by the first distance may be used as it is as the final evaluation.
- Natural sentence selection or rearrangement can be used in combination with natural sentence evaluation 'selection or rearrangement by correlation analysis or scheme analysis, but it is needless to say that it can be used based on the first evaluation value and the second evaluation value or the third evaluation value.
- the mode of evaluating the degree of coincidence between natural sentences is similar to that of evaluating the degree of coincidence using only the number of matching words between the original sentence and the natural sentence.
- a search mode such as a phrase search mode is set in addition to the normal search mode, and this phrase search mode is selected.
- 1st evaluation value and 2nd evaluation value or 3rd evaluation value Evaluation of the degree of coincidence based on the value ' may be performed to select or sort of a natural sentence.
- the natural sentence is selected (judgment of a translation) by calculating the degree of coincidence. If the natural sentence can not be narrowed down by the degree of coincidence, the natural sentence is selected by correlation analysis ( In the case where appropriate natural sentences can not be narrowed down by correlation analysis, natural sentences are selected by schema analysis (translation judgment). Instead, each natural sentence including the words to be translated extracted by the search means is subjected to the calculation of the degree of coincidence, correlation analysis, and scheme analysis to evaluate each natural sentence. Based on the evaluation results, natural sentences may be selected in accordance with the priorities shown in Table 6 below.
- step 104 it is determined whether or not a natural sentence that is completely matched with the original sentence to be processed by the search in step 102 is the extracted DB power. If the determination is affirmative (step 104), a bilingual sentence of a natural sentence that completely matches the original sentence to be processed is read out and displayed (step 106). Steps 104 and 106 described above are omitted, and the natural sentence L, which matches the original sentence to be processed by the search in step 102, regardless of the extracted power, The processing after step 108 (processing such as the calculation of the degree of coincidence) may be performed unconditionally.
- the accuracy of the translation selection in the present invention depends on the number of natural sentences and translations registered in the translation DB, and the accuracy of the translation selection improves as the number of natural sentences and translations increases. .
- the translated original text and the translated text are read into the bilingual determination device according to the present invention, and the read original text and the translated text are selected as they are or after being selected, and then automatically translated into the bilingual DB as a natural text and a translated text.
- the accuracy of the translation selection in the present invention also depends on the redundancy of the contents of the natural sentence and the translated sentence registered in the bilingual DB.
- the accuracy of the bilingual selection decreases as compared with the number of natural sentences and the parallel sentences registered in the bilingual DB (the size of the bilingual DB). For this reason, in the bilingual determination device according to the present invention, there is a natural sentence and a bilingual sentence having a high similarity in content among individual natural sentences and bilingual sentences registered in the bilingual DB! If the natural sentence pair and the translated sentence pair are found, a function to delete one natural sentence and the translated sentence from the bilingual DB may be provided.
- the above description has been made in connection with an example in which a single translation DB is used to select a translation.
- the present invention is not limited to this.
- the bilingual DB may be divided for each field of the original text to be processed. In this case, it is possible to suppress an increase in the capacity of each bilingual DB, shorten the time required for searching for a natural sentence, and improve the accuracy of selecting a bilingual translation.
- the output of the bilingual translation of the translation target phrase performed via the user power PC10 is described.
- the power is not limited to this.
- the Internet or LAN Local Area Network
- the present invention is applied to a computer 'system 38 in which a client PC 32 and a server' computer 34 are respectively connected to a network 30 such as an HDD and a storage medium 36 such as an HDD for storing a bilingual DB is connected to the server 'computer 34'.
- the server / combiner 34 may determine the translation and reply online.
- the source language text data is transmitted to the server computer 34 via the user S client PC32.
- This can be done by designating the source language text by notifying Sano's computer 34 of the URL of the web page containing the source language text or the text of the source language.
- the server computer 34 judges the translation (sentence) of the specified text by executing the translation judgment processing while accessing the translation DB stored in the storage medium 36 ((2) in FIG. 4). )), And send the determined translation (text) to the client PC 32 to answer the question (see (3) in Fig. 4).
- This configuration is effective for cost reduction particularly when the translation DB has a large capacity, since a plurality of users can share the translation determination function (and the translation DB) of the server computer 34.
- FIG. 5 shows an example of a schematic configuration of a machine translation device 40 to which the present invention is applied.
- an original sentence input section 42 inputs (the text data of) the original sentence to be translated to a translation target phrase selecting section 44 and a bilingual sentence assembling section 46, respectively.
- the translation target phrase selection unit 44 selects a specific phrase in the input original text to be translated as the translation target phrase, and inquires the bilingual determination unit 48 of the selected translation target phrase as a translation target phrase. Repeat while sequentially selecting individual words in the original text to be translated.
- Translation The judging unit 48 is a part corresponding to the bilingual judging device according to the present invention, and executes the bilingual judging process while accessing the bilingual DB stored in the storage medium 50, so that the translated phrase to be matched can be obtained. And outputs the determined bilingual translation to the bilingual sentence assembly unit 46.
- the bilingual sentence assembling section 46 is based on the translation target phrase input from the translation target phrase selecting section 44, the bilingual translation input from the bilingual determination section 48, and the original text to be translated input from the original text inputting section 42. By assembling the translations input from the translation judging unit 48 (rearranging the order as necessary), a translation in the target language is assembled.
- the bilingual sentence assembled by the bilingual sentence assembling unit 46 is output to the bilingual sentence output unit 52, and a bilingual sentence output process such as displaying on a display or the like, recording on a recording medium, and outputting as speech is performed.
- the present invention since the present invention is applied to the determination of the translation of the word to be translated, there is a high possibility that a natural translation as the target language sentence can be obtained from the original text as the translation of the word to be translated. Is obtained, and the bilingual sentence output by the bilingual sentence output unit 52 becomes a natural bilingual sentence as a sentence in the target language. In addition, complicated processing such as part-of-speech determination and syntax analysis is not required, and the processing can be simplified.
- the machine translation device shown in FIG. 5, as shown in FIG. 4 described above in response to an online query for a translated text, the translated text obtained through processing such as I'm going to answer it, not to mention it.
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CA002549769A CA2549769A1 (en) | 2003-12-15 | 2004-10-15 | System, method, and program for identifying the corresponding translation |
US10/582,932 US20070112553A1 (en) | 2003-12-15 | 2004-10-15 | System, method, and program for identifying the corresponding translation |
EP04792480A EP1703419A1 (en) | 2003-12-15 | 2004-10-15 | Translation judgment device, method, and program |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2003-416778 | 2003-12-15 | ||
JP2003416778 | 2003-12-15 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2005059771A1 true WO2005059771A1 (ja) | 2005-06-30 |
Family
ID=34697002
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2004/015263 WO2005059771A1 (ja) | 2003-12-15 | 2004-10-15 | 対訳判断装置、方法及びプログラム |
Country Status (6)
Country | Link |
---|---|
US (1) | US20070112553A1 (ja) |
EP (1) | EP1703419A1 (ja) |
KR (1) | KR20060124632A (ja) |
CN (1) | CN1894688A (ja) |
CA (1) | CA2549769A1 (ja) |
WO (1) | WO2005059771A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8077974B2 (en) | 2006-07-28 | 2011-12-13 | Hewlett-Packard Development Company, L.P. | Compact stylus-based input technique for indic scripts |
JP2016071439A (ja) * | 2014-09-26 | 2016-05-09 | パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカPanasonic Intellectual Property Corporation of America | 翻訳方法及び翻訳システム |
Families Citing this family (63)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060116865A1 (en) | 1999-09-17 | 2006-06-01 | Www.Uniscape.Com | E-services translation utilizing machine translation and translation memory |
US7904595B2 (en) | 2001-01-18 | 2011-03-08 | Sdl International America Incorporated | Globalization management system and method therefor |
US8214196B2 (en) | 2001-07-03 | 2012-07-03 | University Of Southern California | Syntax-based statistical translation model |
WO2004001623A2 (en) | 2002-03-26 | 2003-12-31 | University Of Southern California | Constructing a translation lexicon from comparable, non-parallel corpora |
US8548794B2 (en) | 2003-07-02 | 2013-10-01 | University Of Southern California | Statistical noun phrase translation |
US8566081B2 (en) * | 2004-03-25 | 2013-10-22 | Stanley F. Schoenbach | Method and system providing interpreting and other services from a remote location |
US7983896B2 (en) | 2004-03-05 | 2011-07-19 | SDL Language Technology | In-context exact (ICE) matching |
US8296127B2 (en) | 2004-03-23 | 2012-10-23 | University Of Southern California | Discovery of parallel text portions in comparable collections of corpora and training using comparable texts |
US8666725B2 (en) | 2004-04-16 | 2014-03-04 | University Of Southern California | Selection and use of nonstatistical translation components in a statistical machine translation framework |
JP5452868B2 (ja) | 2004-10-12 | 2014-03-26 | ユニヴァーシティー オブ サザン カリフォルニア | トレーニングおよび復号のためにストリングからツリーへの変換を使うテキスト‐テキスト・アプリケーションのためのトレーニング |
JP2006252049A (ja) * | 2005-03-09 | 2006-09-21 | Fuji Xerox Co Ltd | 翻訳システム、翻訳方法およびプログラム |
US8886517B2 (en) | 2005-06-17 | 2014-11-11 | Language Weaver, Inc. | Trust scoring for language translation systems |
US8676563B2 (en) | 2009-10-01 | 2014-03-18 | Language Weaver, Inc. | Providing human-generated and machine-generated trusted translations |
US10319252B2 (en) | 2005-11-09 | 2019-06-11 | Sdl Inc. | Language capability assessment and training apparatus and techniques |
JP4671898B2 (ja) * | 2006-03-30 | 2011-04-20 | 富士通株式会社 | 音声認識装置、音声認識方法、音声認識プログラム |
US8943080B2 (en) * | 2006-04-07 | 2015-01-27 | University Of Southern California | Systems and methods for identifying parallel documents and sentence fragments in multilingual document collections |
US7372851B2 (en) * | 2006-06-28 | 2008-05-13 | International Business Machines Corporation | Method and system for automatic language negotiation on voice (over IP) calls |
US8886518B1 (en) | 2006-08-07 | 2014-11-11 | Language Weaver, Inc. | System and method for capitalizing machine translated text |
US8249855B2 (en) * | 2006-08-07 | 2012-08-21 | Microsoft Corporation | Identifying parallel bilingual data over a network |
US8433556B2 (en) | 2006-11-02 | 2013-04-30 | University Of Southern California | Semi-supervised training for statistical word alignment |
US9122674B1 (en) | 2006-12-15 | 2015-09-01 | Language Weaver, Inc. | Use of annotations in statistical machine translation |
WO2008080190A1 (en) * | 2007-01-04 | 2008-07-10 | Thinking Solutions Pty Ltd | Linguistic analysis |
US8468149B1 (en) * | 2007-01-26 | 2013-06-18 | Language Weaver, Inc. | Multi-lingual online community |
US8615389B1 (en) | 2007-03-16 | 2013-12-24 | Language Weaver, Inc. | Generation and exploitation of an approximate language model |
US8831928B2 (en) | 2007-04-04 | 2014-09-09 | Language Weaver, Inc. | Customizable machine translation service |
JP5299011B2 (ja) * | 2009-03-25 | 2013-09-25 | セイコーエプソン株式会社 | テープ印刷装置、テープ印刷装置の制御方法及びプログラム |
US20120065957A1 (en) * | 2009-05-08 | 2012-03-15 | Werner Jungblut | Interpersonal communications device and method |
US8990064B2 (en) | 2009-07-28 | 2015-03-24 | Language Weaver, Inc. | Translating documents based on content |
US8380486B2 (en) | 2009-10-01 | 2013-02-19 | Language Weaver, Inc. | Providing machine-generated translations and corresponding trust levels |
US10417646B2 (en) | 2010-03-09 | 2019-09-17 | Sdl Inc. | Predicting the cost associated with translating textual content |
US10657540B2 (en) | 2011-01-29 | 2020-05-19 | Sdl Netherlands B.V. | Systems, methods, and media for web content management |
US9547626B2 (en) | 2011-01-29 | 2017-01-17 | Sdl Plc | Systems, methods, and media for managing ambient adaptability of web applications and web services |
US10580015B2 (en) | 2011-02-25 | 2020-03-03 | Sdl Netherlands B.V. | Systems, methods, and media for executing and optimizing online marketing initiatives |
US10140320B2 (en) | 2011-02-28 | 2018-11-27 | Sdl Inc. | Systems, methods, and media for generating analytical data |
US11003838B2 (en) | 2011-04-18 | 2021-05-11 | Sdl Inc. | Systems and methods for monitoring post translation editing |
US8694303B2 (en) | 2011-06-15 | 2014-04-08 | Language Weaver, Inc. | Systems and methods for tuning parameters in statistical machine translation |
US8855997B2 (en) * | 2011-07-28 | 2014-10-07 | Microsoft Corporation | Linguistic error detection |
US9984054B2 (en) | 2011-08-24 | 2018-05-29 | Sdl Inc. | Web interface including the review and manipulation of a web document and utilizing permission based control |
US8886515B2 (en) | 2011-10-19 | 2014-11-11 | Language Weaver, Inc. | Systems and methods for enhancing machine translation post edit review processes |
US8942973B2 (en) | 2012-03-09 | 2015-01-27 | Language Weaver, Inc. | Content page URL translation |
JP5915326B2 (ja) * | 2012-03-29 | 2016-05-11 | 富士通株式会社 | 機械翻訳装置、機械翻訳方法及び機械翻訳プログラム |
US9773270B2 (en) | 2012-05-11 | 2017-09-26 | Fredhopper B.V. | Method and system for recommending products based on a ranking cocktail |
US10261994B2 (en) | 2012-05-25 | 2019-04-16 | Sdl Inc. | Method and system for automatic management of reputation of translators |
US11386186B2 (en) | 2012-09-14 | 2022-07-12 | Sdl Netherlands B.V. | External content library connector systems and methods |
US10452740B2 (en) | 2012-09-14 | 2019-10-22 | Sdl Netherlands B.V. | External content libraries |
US11308528B2 (en) | 2012-09-14 | 2022-04-19 | Sdl Netherlands B.V. | Blueprinting of multimedia assets |
US9916306B2 (en) | 2012-10-19 | 2018-03-13 | Sdl Inc. | Statistical linguistic analysis of source content |
US9152622B2 (en) | 2012-11-26 | 2015-10-06 | Language Weaver, Inc. | Personalized machine translation via online adaptation |
JP6417649B2 (ja) * | 2013-08-22 | 2018-11-07 | 株式会社リコー | 文章処理装置、文章表示システム、プログラム |
US9213694B2 (en) | 2013-10-10 | 2015-12-15 | Language Weaver, Inc. | Efficient online domain adaptation |
US9734820B2 (en) * | 2013-11-14 | 2017-08-15 | Nuance Communications, Inc. | System and method for translating real-time speech using segmentation based on conjunction locations |
JP6074820B2 (ja) * | 2015-01-23 | 2017-02-08 | 国立研究開発法人情報通信研究機構 | アノテーション補助装置及びそのためのコンピュータプログラム |
KR101709693B1 (ko) * | 2015-07-13 | 2017-02-27 | 주식회사 에버트란 | 크라우드 소싱을 이용한 웹툰 언어 자동 번역 방법 |
US10614167B2 (en) | 2015-10-30 | 2020-04-07 | Sdl Plc | Translation review workflow systems and methods |
US11049190B2 (en) | 2016-07-15 | 2021-06-29 | Intuit Inc. | System and method for automatically generating calculations for fields in compliance forms |
US10579721B2 (en) | 2016-07-15 | 2020-03-03 | Intuit Inc. | Lean parsing: a natural language processing system and method for parsing domain-specific languages |
JP7117629B2 (ja) * | 2017-04-27 | 2022-08-15 | パナソニックIpマネジメント株式会社 | 翻訳装置 |
US10635863B2 (en) | 2017-10-30 | 2020-04-28 | Sdl Inc. | Fragment recall and adaptive automated translation |
US10817676B2 (en) | 2017-12-27 | 2020-10-27 | Sdl Inc. | Intelligent routing services and systems |
US11256867B2 (en) | 2018-10-09 | 2022-02-22 | Sdl Inc. | Systems and methods of machine learning for digital assets and message creation |
US10831999B2 (en) * | 2019-02-26 | 2020-11-10 | International Business Machines Corporation | Translation of ticket for resolution |
US11163956B1 (en) | 2019-05-23 | 2021-11-02 | Intuit Inc. | System and method for recognizing domain specific named entities using domain specific word embeddings |
US11783128B2 (en) | 2020-02-19 | 2023-10-10 | Intuit Inc. | Financial document text conversion to computer readable operations |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH04160473A (ja) * | 1990-10-24 | 1992-06-03 | Hitachi Ltd | 事例再利用型翻訳方法および装置 |
JP2000020524A (ja) * | 1998-07-02 | 2000-01-21 | Fuji Xerox Co Ltd | 対訳文検索装置 |
JP2003263434A (ja) * | 2002-03-11 | 2003-09-19 | Advanced Telecommunication Research Institute International | 翻訳システムの自動選択をコンピュータに実行させるためのプログラム、およびそのプログラムを記録したコンピュータ読取り可能な記録媒体 |
Family Cites Families (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE68928231T2 (de) * | 1988-10-28 | 1998-01-08 | Toshiba Kawasaki Kk | Verfahren und Vorrichtung zur Maschinenübersetzung |
US5497319A (en) * | 1990-12-31 | 1996-03-05 | Trans-Link International Corp. | Machine translation and telecommunications system |
JPH09128396A (ja) * | 1995-11-06 | 1997-05-16 | Hitachi Ltd | 対訳辞書作成方法 |
US5983169A (en) * | 1995-11-13 | 1999-11-09 | Japan Science And Technology Corporation | Method for automated translation of conjunctive phrases in natural languages |
US6415250B1 (en) * | 1997-06-18 | 2002-07-02 | Novell, Inc. | System and method for identifying language using morphologically-based techniques |
US6192332B1 (en) * | 1998-04-06 | 2001-02-20 | Mitsubishi Electric Research Laboratories, Inc. | Adaptive electronic phrase book |
US6345244B1 (en) * | 1998-05-27 | 2002-02-05 | Lionbridge Technologies, Inc. | System, method, and product for dynamically aligning translations in a translation-memory system |
JP3879321B2 (ja) * | 1998-12-17 | 2007-02-14 | 富士ゼロックス株式会社 | 文書要約装置、文書要約方法及び文書要約プログラムを記録した記録媒体 |
JP3822990B2 (ja) * | 1999-01-07 | 2006-09-20 | 株式会社日立製作所 | 翻訳装置、記録媒体 |
US6243669B1 (en) * | 1999-01-29 | 2001-06-05 | Sony Corporation | Method and apparatus for providing syntactic analysis and data structure for translation knowledge in example-based language translation |
US6438524B1 (en) * | 1999-11-23 | 2002-08-20 | Qualcomm, Incorporated | Method and apparatus for a voice controlled foreign language translation device |
US6473729B1 (en) * | 1999-12-20 | 2002-10-29 | Xerox Corporation | Word phrase translation using a phrase index |
US7054803B2 (en) * | 2000-12-19 | 2006-05-30 | Xerox Corporation | Extracting sentence translations from translated documents |
US7191115B2 (en) * | 2001-06-20 | 2007-03-13 | Microsoft Corporation | Statistical method and apparatus for learning translation relationships among words |
US7353165B2 (en) * | 2002-06-28 | 2008-04-01 | Microsoft Corporation | Example based machine translation system |
US7346487B2 (en) * | 2003-07-23 | 2008-03-18 | Microsoft Corporation | Method and apparatus for identifying translations |
GB2415518A (en) * | 2004-06-24 | 2005-12-28 | Sharp Kk | Method and apparatus for translation based on a repository of existing translations |
US7672830B2 (en) * | 2005-02-22 | 2010-03-02 | Xerox Corporation | Apparatus and methods for aligning words in bilingual sentences |
-
2004
- 2004-10-15 CN CNA2004800374589A patent/CN1894688A/zh active Pending
- 2004-10-15 CA CA002549769A patent/CA2549769A1/en not_active Abandoned
- 2004-10-15 US US10/582,932 patent/US20070112553A1/en not_active Abandoned
- 2004-10-15 WO PCT/JP2004/015263 patent/WO2005059771A1/ja active Application Filing
- 2004-10-15 EP EP04792480A patent/EP1703419A1/en not_active Withdrawn
- 2004-10-15 KR KR1020067011763A patent/KR20060124632A/ko not_active Application Discontinuation
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH04160473A (ja) * | 1990-10-24 | 1992-06-03 | Hitachi Ltd | 事例再利用型翻訳方法および装置 |
JP2000020524A (ja) * | 1998-07-02 | 2000-01-21 | Fuji Xerox Co Ltd | 対訳文検索装置 |
JP2003263434A (ja) * | 2002-03-11 | 2003-09-19 | Advanced Telecommunication Research Institute International | 翻訳システムの自動選択をコンピュータに実行させるためのプログラム、およびそのプログラムを記録したコンピュータ読取り可能な記録媒体 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8077974B2 (en) | 2006-07-28 | 2011-12-13 | Hewlett-Packard Development Company, L.P. | Compact stylus-based input technique for indic scripts |
JP2016071439A (ja) * | 2014-09-26 | 2016-05-09 | パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカPanasonic Intellectual Property Corporation of America | 翻訳方法及び翻訳システム |
Also Published As
Publication number | Publication date |
---|---|
CA2549769A1 (en) | 2005-06-30 |
US20070112553A1 (en) | 2007-05-17 |
CN1894688A (zh) | 2007-01-10 |
KR20060124632A (ko) | 2006-12-05 |
EP1703419A1 (en) | 2006-09-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2005059771A1 (ja) | 対訳判断装置、方法及びプログラム | |
JP5235344B2 (ja) | 機械翻訳を行う装置、方法およびプログラム | |
US20060241934A1 (en) | Apparatus and method for translating Japanese into Chinese, and computer program product therefor | |
US8812296B2 (en) | Method and system for natural language dictionary generation | |
JP2007257644A (ja) | 訳語候補文字列予測に基づく訳語取得のためのプログラム、方法および装置 | |
JP2010157178A (ja) | テキスト・データに含まれる固有表現又は専門用語から用語辞書を作成するためのコンピュータ・システム、並びにその方法及びコンピュータ・プログラム | |
JP2006252382A (ja) | 質問応答システム、およびデータ検索方法、並びにコンピュータ・プログラム | |
JP2007241764A (ja) | 構文解析プログラム、構文解析方法、構文解析装置、及び構文解析プログラムが記録されたコンピュータ読み取り可能な記録媒体 | |
Yalcin et al. | An external plagiarism detection system based on part-of-speech (POS) tag n-grams and word embedding | |
JP2020190970A (ja) | 文書処理装置およびその方法、プログラム | |
JPWO2010061733A1 (ja) | 誤訳の検出を支援する装置及び方法 | |
US8554539B2 (en) | Method for analyzing morpheme using additional information and morpheme analyzer for executing the method | |
JP4640593B2 (ja) | 多言語文書検索装置および多言語文書検索方法、並びに、多言語文書を検索するプログラム | |
JP4401269B2 (ja) | 対訳判断装置及びプログラム | |
CN111767733A (zh) | 一种基于统计分词的文献密级甄别方法 | |
US10572592B2 (en) | Method, device, and computer program for providing a definition or a translation of a word belonging to a sentence as a function of neighbouring words and of databases | |
Ullah et al. | Opinion summarization of bangla texts using cosine simillarity based graph ranking and relevance based approach | |
Nghiem et al. | Using MathML parallel markup corpora for semantic enrichment of mathematical expressions | |
JP2003323425A (ja) | 対訳辞書作成装置、翻訳装置、対訳辞書作成プログラム、および翻訳プログラム | |
JP2005326952A (ja) | 概念辞書への単語登録方法、装置、およびプログラム | |
JP4985096B2 (ja) | 文書解析システム、および文書解析方法、並びにコンピュータ・プログラム | |
JP2000250914A (ja) | 機械翻訳方法、装置、および機械翻訳プログラムを記録した記録媒体 | |
JP3419748B2 (ja) | 辞書作成装置および方法と辞書作成プログラムを記録した記録媒体 | |
Bhowmik et al. | Development of A Word Based Spell Checker for Bangla Language | |
JP4262529B2 (ja) | 全文検索装置、方法、プログラム及び記録媒体 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 200480037458.9 Country of ref document: CN |
|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DPEN | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed from 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2007112553 Country of ref document: US Ref document number: 1020067011763 Country of ref document: KR Ref document number: 10582932 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2549769 Country of ref document: CA |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2004792480 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 2004792480 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1020067011763 Country of ref document: KR |
|
WWP | Wipo information: published in national office |
Ref document number: 10582932 Country of ref document: US |