CA2120447C

CA2120447C - Automatically relaxable query for information retrieval

Info

Publication number: CA2120447C
Application number: CA002120447A
Authority: CA
Inventors: Robert Lizee; Martin Lizee
Original assignee: Individual
Current assignee: Individual
Priority date: 1994-03-31
Filing date: 1994-03-31
Publication date: 1998-08-25
Anticipated expiration: 2014-03-31
Also published as: CA2120447A1; US5671404A

Abstract

A data processing system is disclosed for automatically querying databases. The system uses an Automatically Relaxable Query (ARQ) consisting of a list of query conditions ordered in descending degree of importance. When an ARQ is processed, the database is queried on the first condition of the ARQ; if any object is found, the first condition is labelled. Then, the system proceeds with a query composed of all the conditions labelled previously in conjunction with the next condition of the ARQ; if any object is found then this next condition is labelled. The last step is repeated until all the remaining conditions of the ARQ have been used. The system then reports which conditions were labelled. If the query composed of the labelled conditions matches only few objects of the database, the objects may be automatically retrieved. The automatic querying with an ARQ enables faster searches, minimizes interaction between the database and the user and permits off-line preparation of database querying.

Description

212~4~17 ~y~-~ r ~ FO~ ~U~YlN~ ~A~A~A~ES AUTOMATICALLY

SPECIFICATION

FIELD OF THE lNvL...lON

The invention di~closed herein relates to the field of information retrieval, and particularly to the ; ~_~ved perf~ -n~e of information retrieval through the use of - ~-.~ , automatic query processing. The invention may be used in any setting where information retrieval is desired. Such ~rpl;cAtions include, but are not limited to, text databases, relational datAhA~es, and off-line querying of datAhA~es.
. .. : . , .:; .:
.: .~: ~.' ' ''.'.,.

RA~OUND OF T~E INVENTION

Information retrieval involves the search of retrieval objects in database systems through the use of information retrieval queries followed by the ou~uLLing of the retrieval objects found. Records and do~l ~nts are typical retrieval obiects.
~ ~ :
A major issue in information retrieval is how to query database systems effectively in order to find the desired retrieval objects. Generally, the way to find retrieval objects is for the ~ ;
user to perform queries interactively until the desired retrieval objects are found. This search involves a continuous interaction bet.~een the user and the database system by trial and error while searching for the desired retrieval objects.
" .' ' ,,:' "'., ' .,.

:: ~-'.' ' ' , ' :, ' '~ ' ' ~: .~:
!

212~7 In interactive querying, users con~tantly evaluate the results ~ -of a query and requery the database until the desired retrieval objects are found. Although this method may give excellent results the process of querying is time consuming, eYpen~ive and repetitive.

A database system con~A;n; ng a large quantity of retrieval objects is more complex to search becAll~e retrieval objects are ~-more likely to share common characteristics. A user must be more spec; f ic when querying the database system in order to find the desired retrieval objects, since otherwise the number of retrieval objects found would be too large to be useful. This i8 true for large relational databases system and is ss~ec;Ally - -~
true for text datAhA~e~ system wherein most of the words are Y~ " ~;
,, ' ~, ~':~ ' Most text datAhA~e systems are also complex to guery becAl~e they are not well-structured and the user cannot rely on the st~u~u~e of the database for querying.

Another difficulty that database users have to overcome is to ~gg;gn effective queries to find the desired retrieval objects.
Nost query languages require a good knowledge of boolean logic and of the database by the user.

, '.
An attempt to overcome the drAwhAc~ of interactive guerying by using a quantitative retrieval method is given in US Patent Number 5,263,159 issued on Nov. 16, 1993. In that method, the user asks the database to find the documents that best match the conditions of the query without necessarily matching all these ~ --conditions. Each condition of the query ha~ a weight and the

2 ';~ '~'''' ' ~ ';

~: ~,'' '"

2120~47 score for a retrieval object of the database consist of the sum of the weights of all the conditions sati~fied by the retrieval object. Only do~, -nts with the best scoree are brought to the attention of the user.

The quantitative retrieval methods, using a weight factor for each condition, have the disavantage in that the user must select a weight for each condition. Selection of the appropriate weight factor i~ not intuitive and therefore not always an easy cho;ce to make.

., : , Furth~ e the search results do not provide sufficient conf;~ s to the user that all u~eful retrieval objects have been found hecA~ e the evolution of the search p~ocess iB not tr~n~pArent to the user.

Finally, quantitative retrieval can be a long ~Loce~s since, for a number of N conditions in a quantitative retrieval guery, up to 2N qusries might be .ee~e~ to find the desired retrieval ~--ob~ects. ~ 1 ~.., ' In practice, a user searching interactively in a database system ~;
adds conditions to a query in order to pin-point a set of ;~
relevant retrieval objects- The ~eC; ~i on to add a new query condition often ~p~ only on the number of retrieval objects found rather than the contents of the retrieval object found - - -a large number of f;n~;ngs may necessitate further conditions.
: ' . '', , '::
"".,.",.......
,~ ~,', .'' ' . ,' SUMMARY OP THE INVENTION

3 --21204~7 An a~pect of the pre~ent invention is to overcome the problems described above by establishing a query system of an ordered list of all po~sible query conditions for the search, ordered from the most important to the least important query condition which forms the 80 called Automatically Re~ hle Query (ARQ).
'; ~ ',,.,'.

It is a further aspect of the invention to automate the query p~Gcess 80 that after defining an ARQ and l~nnch;ng it, the user only waits for the results. The process will query automatically the datAhA~e system, rslAY;ng the ARQ if neeAe~ to find the desired retrieval objects. An advantage of this query system is that it i8 less likely for human mistakes to occur since the user has only to ~f;ns an ARQ and doesn't have to directly query the database. It also permits off-line searching for remote datAhA~e systems. The user can therefore define ARQs without having is computer co~n~cted to the database system, and when con-eL-~ed to it, no user intervention is ~eeAe~ since the ARQ will automatically do the search- C~ n;~Ation with the datahA~e system is hereby optimized since there is no 'dead time', in the sense that there are no moments where a user needs to think about the search while the computer i8 co~.neu~ed to the database.

A further advantage of the pla~ellt query ~loce88 is greater speed since the number of queries that the process will have to do when querying the database with the ARQ is linear with the number of condition~ ~sf;n;ng the ARQ. The querying processing speed is im~ uved since machine/machine interaction is faster than a machine/human interaction.

The ~ eser,t query system saves communication time, and all the . ..
co~ts that can be related to it heCAll~e the query pLocess is faster and more optimized, enabling a user to find desired retrieval objects with only a few ARQ. It also enables databases . . :
to accommodate more users with the same resources since the connections are optimized, and users are more satisfied with the result of their search and the cost related to it.
'','., The present invention can be used on any database system without modification to the database system itself, just by. Furthe -re an ARQ can be defined in a general way without being depe~Ant on a particlllAr database system. The query processing of an ARQ
is adaptable to different database systems. Therefore, since an ARQ is proces~e~ automatically, it is possible to query different database systems automatically with the same ARQ.

An aspect of the invention is in the dete ;n;~tic nature of the querying process whe,cb~ a query condition of the ARQ is not marked if and only if the subset of conditions that contains it , . ... ...
together with the marked conditions of higher rank in the ARQ ~-satisfies no retrieval objects (or the marked conditions of ~; -. . . ~ .
h; gher ranks already lead to a sufficiently low number of ~o _ Ls). Knowing this, a user can easily understand why a condition was not marked.
~. , : .:
,~ ,- . .:. - . :
- . . . i,.: ~, ~ :,:..
AccoLding to the p~eser.L invention it is ~ ose~l to ~L~3ent the results of the ARQ search in a simple and intuitive way by - .,: .: :
informing the user of the marked conditions of the ARQ, and by -- ;~
.. .. ~.
showing the num~er of retrieval objects found for each :.. : ~ .--:
combination of query conditions. ~ ~ ~
. . . .

A further aspect of the invention is an automatic information ~ - ~
,- ,~ ': ~:' " ' '.,. - ":
.

,.' ' ~.....

212~7 -~

retrieval method for use with a computer having a link to a ~ -database system containing a plurality of retrieval objects, each retrieval object satisfying one or more query conditions, a query inputting device for specifying a query expressed as a conjunction of query conditions, the information retrieval method comprising the steps of: -(a) inputting a list of at least two query conditions ordered from the most important query condition to the least important query condition wherein each query condition of the list can be marked and starts as being unmarked;
(b) designating as the current query condition the most important query condition of said list of query conditions;
(c) querying said database system for retrieval objects, using said query inputting device, with a query composed of ~aid current query condition;
(d) marking said current query condition if retrieval objects are ~ound in step (c);
(e) ~ss;q~ting as current query condition the next less ;
important query condition in said list; -;~
(f) querying said database ~ystem for retrieval objects, using ~ '' said query inputting device, with a query composed of the :~
conjunction of all previously marked query conditions and said current que~y condition;~
(g) marking ~aid current query condition if retrieval objects are found in step (f); ~ ::
(h) repeating steps (e), (f) and (g) until all query conditions of said list have been designated once as current :~
query condition.

BRIBF ~C~RTPTION OF T~E DRAWINGS

21204~7 :
; . . .
Figure lA is a flowchart showing the basic ARQ process.
Figure lB is a flowchart showing the extended ARQ process.
Figure 2 is an example of the execution of an ARQ.
Figure 3 is an illustration showing some possible communication links between a computer and a remote database system.

DE~ATTRn DR.5~PTPTION OF THE lNV~n ~ lON

Before pLocee~;ng to the drawings, the nature of the queries will be described.

A database system contains a plurality of retrieval objects. The - - ;
datAhA~es system contains a query inputting device for searching ; ;
,;, -for desired retrieval objects. The object of the query inputting -device is to input queries, search the database system, and to ~ :~
return the result of those gueries in the form of the number of retrieval objects found and/or the objects found.

If the datAhA~e system is local to the computer, it is assumed here to be l;nks~ to it, whereas for a remote database a link must be establ;~he~. As shown in figure 3, this link can be - -estAhl; Yh~ with - - via a phone line, through local area networks or through any cther means possible.

A query condition as herein defined is any boolean condition ;
which ia used for querying retrieval objects. If a retrieval object respects a query condition, it responds positively to it, ~;
raturning value true, otherwise it re~ponds negatively. A query condition can be c~ -~e~ of many query conditions using usual boolean connectives. For example, "A or B", provided that "A" - ~
: ~' . ,-. ',::
7 ~

212~7 and "B" are query condition~.

A query is defined here as a conjunction of query conditions.
The query ~earches for c n retrieval objects respo~; ng positively to all its query conditions.

In the case of a full text database system, a retrieval object co,Leepo~ to a document or parts of it, a query condition is the ~Lc3ence of a specific word ~e~L~ssion) or its equivalent word ex~Lession in a do~ t and a query is a list of those query conditions. Here we say equivalent instead of synonymous hecAn~e, ~epe~; ng on the context of the search, a user might cQ~q;~sr two words that are not synGr,~ - ~ as being eguivalent.
".,... ~.:
For instance, divorce might be eguivalent to separation. A word ~-here can also be an expres~;Q~ like "family law". -In the case of a relational database system, a retrieval object cGLL~_, to a L~coLd, a query condition coLL~s~onds to an SQL
condition and a query coLL-~l,on~ to an SQL statement made of -~
it~ SQL conditions l; n~.e~ with the SQL operator "A~D~
~: -",- ~ ':,:.
Figures lA and lB are flow charts illustrating the logical ~ ~ -operations within the c~ ~er. Figure lA shows'the basic ARQ
pLoce~8 and figure lB shows the extende~ ARQ pLocess comprising ~-possible options. ~ ;

- Input an Automatically RelAY~hle Query (box 100) The method must first input a list of conditions ordered from the most important query condition to the least important query condition. This list is called the Automatically RelA~Ahle Query ~ . ' (ARQ). The user chooses the relative importance of a query condition. The user chooses the most relevant query condition as being the most important and 80 on. An ARQ must have at least two query conditions, otherwise it would be a standard query.
The ARQ may be inputted from the user or from a file or any other form of input.
~ .. ... .
Each query condition of the ARQ is marked or unmarked. All query conditions start out as being unmarked.
','.';' ':','' ~' - Co~n~ction to the database (box 105) In the case of a remote database system, the link may be made ~ ; ;
and broken as part of the querying. Here, a cQnnsction is ~ -... . :, .
estAhl;~he~ bet~ - the computer and the database system through ~ - -the link. The conneation includes whatever steps are necessary to attain the mode which makes querying of the database system possible.

- Automatically querying database system with ARQ (boxes 120 to 190 ; ncl n~l~l ) . ' : . , ~
:, ' , . I i ! : .
The ~_ ~er queries the database system automatically with the ARQ. To do 80, the computer must first query the database for -retrieval objects that respond po~itively to the first query condition of the ARQ (boxes 110, 120, 130). If at lea~t one ~ ; -retrieval object is found, the computer marks the first query condition (140). If no objects are found, the computer leaves the first q-aery condition unmarked. The computer then queries the datAh~e system for retrieval objects that respond positively to the seco~ query condition (box 150) and all of ~ .:
: - :
, . . .

2~2~7 the previously marked query conditions (box 160). If at lea~t one retrieval object i8 found (box 170), the computer marks the second condition (box 180). If no object~ are found, the computer leaves the second query condition unmarked. The o~ -Ler repeats the ~ame sequence with the remaining query ~ conditions (boxes 150, 160, 170, 180, 190). Such a procedure may be as follows~
(a) Deeignate the first condition a~ the current query . ~ ~ :
condition (box 110); .
(b) Query the database for retrieval objects with a query composed of the current guery condition (box 120); -.
(c) Mark the current query condition if retrieval objects are ~ ~ ~
found in (b) ~boxes 130 and 140); : ~-(d) Dee;g~Ate as current condition the next less important ~.
query condition (box 150);
(e) Query the databaee system for retrieval objects with a ~-guery composed with the conjunction of all previously marked query conditions and the current query condition (box 160); : :
(f) Mark the ~Ler.~ query condition if retrieval objects are found in (e) (boxes 170 and 180);
(g) Repeat steps (d), (e) and (f) until all query conditions of the ARQ have been designated once as current (box 190). :-An example of the Appl;~Ation of this procedure i~ presented in figure 2, wherein Cl Le~Lesents the first condition of the ARQ, C2 the secQ~, and 80 on. The querie~ are presented as a conjunction of conditiDns l;nk~ by the symbol '&'. -~
:';; :.''':~
It is also possible to te ;nAte the querying when a ~Lede~ermined number of retrieval objects has been found. For instance, if N represents the lower limit of retrieval objects " ~ , " ",,. ;, ~,,, ;

212~4~7 - ,, to be found (box 101), ~tep (g) i8 modified to read:
(g) Repeat (d), (e) and (f) until all conditions from the ARQ
have been designated, or that the number of retrieval objects -~
found in (e) is less than or equal to N (box 191).
'. :', - Retrieve retrieval objects, if desired (boxes 200, 201 and 202) ;.' ,:' ~:' ' ' In this step the computer queries the database system for - -~
retrieval objects that satisfy all the selected query conditions of the ARQ by using a query composed of all these selected query conditions (box 201), and then retrieves the retrieval objects found ~box 202). Only parts of the retrieval objects found may be retrieved if the database system supports partial retrieval of its retrieval objects. For instance, in a text database syatem, only the titles of the documents found may be retrieved.

It is also possible to make this step conditional (box 200) -; . .: ~:. :, guerying and retrieving if the number of retrieval objects found ~; ~ 18 not greater than a previously chosqn -r of desired ~-retrieval objects. -- D;~aQ~ne-~ion from database system (box 210) -: , ~ ::"-:
This step is done in the case of a remote database system. It is also poss~hle to ~oaess other specified ARQ's before A~onnecting from databa~e system. This step is important in -~
order to make the co~nection time as short as possible.

- Show results (box 220) ;

21204~7 In this optional step the computer generates a search report.
The report may list the conditions of the ARQ, specifying which of those were marked and may also show retrieval objects.
~,,~'~ ;'',''''' - Other It is also possible to implement the method directly in the da~AhA~e server, the database system would then directly ~U~O
ARQ and query automatically its own database, instead of having the computer to automatically query the system. A good use of this alternative is in electronic mail where a user would simply :, .
mail an ARQ to a database system which would treat the ARQ and e~uL.. the result to the user.

Another advantAgeon~ way of p.eaen~ing the treatment of an ARQ
is as a systematic reduction of a set of retrieval objects. The treatment of an ARQ would then COL~e~PO~ to the following~
~ a) ~ss;qr~te the first condition as the current query condition;
b) find the retrieval objects es~on~;nq positively to current query condition in the set of retrieval objects;
(c) reduce the set of retrieval objects to only those that are ~ -found in step ~b) if some retrieval objects are found in step ~ ;~
(b); -d) Ass;qnAte as ~uLLen~ condition the next less important query condition in the list of query conditions; ~ -~ e) find the retrieval objects respo~;nq positively to current query condition in the set of retrieval objects;
~f) ,cduce the set of retrieval objects to only those that are ' found in ste2 ~f) if some retrieval objects are found in step ~e);

f- 12 , l :, l : .

212~447 --(g) repeat steps (d), (e) and (f) until all query conditions of the list of query conditions have been designated oncc as ~ .
current query condition.

Although a specific ~ t of the invention has been disclosed, it will be understood by those having skill in the ~:
art that the foregoing and other changes can be made to the ~pecific embodiment without departing from the spirit and the scope of the invention. .~ ~

:,, .

::

. ;'', ' '. -, "'

Claims

Claims:

1. A method for automatically retrieving information from a database through a computer in communication with the database, the database having a plurality of retrieval objects, the computer defining a plurality of pre-ordered and distinct query conditions, wherein each query condition is ordered in a query condition list from the most important query condition to the least important query condition and wherein the most important query condition of the query condition list is initially designated as the current query condition, comprising the steps of:
a) testing the current query condition within the database to determine the number of database retrieval objects which satisfy the current query condition;
b) marking said current query condition if at least one database retrieval object satisfies the current query condition and designating the next pre-ordered and distinct query condition in the query condition list as the current query condition;
c) successively repeating steps a) and b) with each pre-ordered and distinct query condition of the query condition list in conjunction with all previously marked query conditions until a predetermined termination condition is satisfied.

2. A method as in claim 1 wherein after said predetermined termination condition is satisfied, step c) further comprises i) reporting the number of database retrieval objects which satisfy the query condition for each query condition; and, ii) identifying said retrieval objects to the computer.

3. The method as in claim 2 wherein the pre-determined termination condition requires that all query conditions have been the current query condition or that a pre-set minimum number of database retrieval objects which satisfy the query condition is determined.

4. The method as in claim 3 wherein the database system is a relational database system.

5. The method as in claim 4 wherein the query condition is a structured query language (SQL) statement.

6. The method as in claim 3 wherein the database system is a full text database system and the retrieval objects are documents.

7. The method as in claim 6 wherein the query condition is a word or word group and the testing of step a) is satisfied if a database document includes said word or word group.

8. The method as in claim 7 wherein the computer is remote to the database and is connected to the database prior to the initiation of step a) and disconnected from the database after said termination condition is achieved.

9. The method as in claim 1 wherein the pre-determined termination condition requires that all query conditions have been the current query condition.

10. The method as in claim 1 wherein the pre-determined termination condition requires that all query conditions have been the current query condition or a pre-set minimum number of database retrieval objects found in step a) is determined.

11. The method as in claim 1 wherein step c) is continued while a pre-set maximum number is less than the number of database retrieval objects found in step a).

12. The method as in claim 1 wherein the database system is a relational database system.

13. The method as in claim 1 wherein the query condition is structured query language (SQL) statement.

14. The method as in claim 1 wherein the database system is a full text database system and the retrieval objects are documents.

15. The method as in claim 1 wherein the query condition is a word or word group and the testing of step a) is satisfied if a database document includes said word or word group.

16. The method as in claim 1 wherein the computer is remote to the database and is connected to the database prior to the initiation of step a) and disconnected from the database after said termination condition is achieved.

17. A method for retrieving information from a database by an automatically relaxable query through a computer having a link with the database, the database having a plurality of retrieval objects, each retrieval object satisfying one or more query conditions, the automatically relaxable query being defined in the computer as a list of at least two query conditions ordered from the most important query condition to the least important query condition wherein each query condition of the list can be marked and starts as being unmarked and where the first query condition starts as being designated as the current query condition, comprising the steps of:
(a) testing the database for retrieval objects, with a query composed of the conjunction of all marked query conditions and the current query condition if at least one query condition of the list is marked, or with a query composed of the current query condition if there is no query condition marked in the list;
(b) marking the current query condition if retrieval objects are found in step (a);
(c) repeating steps (a) and (b) with the next query condition of the list designated as the current query condition until a predetermined termination condition is satisfied.

18. An automatic information retrieval system for use with a computer having a link to a database system containing a plurality of retrieval objects, each retrieval object satisfying one or more query conditions, a query inputting device for specifying a query expressed as a conjunction of query conditions, the information retrieval system comprising:
(a) means for inputting a list of at least two query conditions ordered from the most important query condition to the least important query condition wherein each query condition of the list can be marked and starts as being unmarked;
(b) means for designating as the current query condition the first query condition of said list of query conditions;
(c) means for querying said database system for retrieval objects, using said query inputting device, with a query composed of said current query condition;
(d) means for marking said current query condition if retrieval objects are found in (c);
(e) means for designating as current query condition the next query condition in said list of query conditions;
(f) means for querying said database system for retrieval objects, using said query inputting device, with a query composed of the conjunction of all previously marked query conditions and said current query condition;
(g) means for marking said current query condition if retrieval objects are found in (f);

(h) means for repeating (e), (f) and (g) until all query conditions of said list have been designated once as current query condition.

19. An automatic information retrieval method for use with a computer having a link to a database system containing a plurality of retrieval objects, each retrieval object satisfying one or more query conditions, a query inputting device for specifying a query expressed as a conjunction of query conditions, the information retrieval method comprising the steps of:
(a) inputting a list of at least two query conditions ordered from the most important query condition to the least important query condition wherein each query condition of the list can be marked and starts as being unmarked;
(b) designating as the current query condition the first query condition of said list of query conditions;
(c) querying said database system for retrieval objects, using said query inputting device, with a query composed of said current query condition;
(d) marking said current query condition if retrieval objects are found in step (c);
(e) designating as current query condition the next query condition in said list of query conditions;
(f) querying said database system for retrieval objects, using said query inputting device, with a query composed of the conjunction of all previously marked query conditions and said current query condition;

(g) marking said current query condition if retrieval objects are found in step (f);
(h) repeating steps (e), (f) and (g) until all query conditions of said list have been designated once as current query condition.