Document retrieval system involving ranking of documents in accordance with a degree to which the documents fulfill a retrieval condition corresponding to a user entry

• US 5,535,382 A
• Filed: 11/17/1993
• Issued: 07/09/1996
• Est. Priority Date: 07/31/1989
• Status: Expired due to Term

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Ogawa, Yasushi
• Primary Examiner(s): Black, Thomas G.
• Assistant Examiner(s): Choules, Jack M.
• Application Number:
  US08/153,279
• Time in Patent Office: 965 Days
• Field of Search: 395/600, 395/800, 395/934, 395/900, 364/419.08, 364/419.19, 364/225.4, 364/274.6, 364/282.1, 364/963, 364/963.1, 364/974, 364/974.6
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: Y10S 706/934 Information retrieval or Information management
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A document retrieval system retrieves one or a plurality of registered documents from a document database responsive to retrieval conditions designated by a user. The document retrieval system includes a query converter for converting the retrieval condition designated by the user into a query which has a predetermined normal form in which keywords and at least one type of logical operation out of logical operations AND, OR and NOT are connected, a bibliographical information indicator for indicating a relation between each of said registered documents and keywords and a keyword connection table having relationship values, each of the relationship values representing the degree of relationship between each two keywords. The document retrieval system also includes a selector for referring the inverted file and the keyword connection so as to select one or a plurality of registered documents which satisfy the query, and an outputting circuit for outputting one or a plurality of registered documents selected by the selecting means.

Claims

1. A document retrieval system for retrieving documents from documents registered in a document database responsive to a retrieval condition designated by a user, said document retrieval system comprising:

a) query converting means for converting said retrieval condition designated by the user into a query having a predetermined normal form in which keywords and at least one type of logical operation out of logical operations AND, OR and NOT are connected;

b) bibliographical information indicating means for indicating at least information concerning which keywords correspond to each document of the documents registered in said document database;

c) a keyword connection table having relationship values, each of the relationship values representing a degree of relationship between keywords;

d) ranking means for ranking documents in accordance with relevance values, each relevance value indicating a degree to which a document fulfills the retrieval condition corresponding to the query, each of said relevance values being calculated for the document using the relationship values provided in said keyword connection table, wherein connected keywords are obtained with reference to the information indicated by said bibliographical indicating means as (1) keywords corresponding to the document and (2) keywords included in the query obtained by said query converting means;

e) outputting means for outputting, as a retrieval result, the documents ranked by said ranking means;

f) inputting means for inputting evaluation information indicating a degree to which each of the documents output by said outputting means is relevant to a document required by the user; and

g) learning means for modifying one or a plurality of said relationship values in said keyword connection table based on the evaluation information input by said inputting means;

wherein said query converting means includes first means for generating a first query in a conjunctive normal form, said first query in the conjunctive normal form having subqueries which are connected with each other only by logical AND operations, each of said subqueries being expressed by at least one keyword connected by at least one of logical OR and NOT operations; and

wherein said ranking means includes;

1) first calculation means for calculating a sub-relevance value for each of said subqueries, said sub-relevance value indicating a degree to which each document fulfills each of said subqueries; and

2) second calculation means for calculating a relevance value for each of the documents using said sub-relevance value calculated for each of said subqueries by said first calculation means.

2. A document retrieval system as claimed in claim 1, wherein said second calculation means calculates said relevance value in accordance with the following formula, ##EQU33## where ri is said relevance value, ri(h) is said sub-relevance value calculated by said first calculation means and N is the number of said subqueries making up said query.

5. A document retrieval system as claimed in claim 1 or claim 3, wherein said ranking means includes:

sorting means for sorting the documents in a downward sequential order of said relevance values corresponding to the documents.

6. A document retrieval system as claimed in claim 5, wherein said relevance value is a value within a range which is equal to or greater than 0 and is equal to or less than 1.

7. A document retrieval system as claimed in claim 1 or claim 3, wherein:

said evaluation information has an evaluation value which is a value indicated by a real number; and

the evaluation value corresponds to the degree of relevance required by the user for each of the documents.

8. A document retrieval system as claimed in claim 7, wherein said relevance value is a value within a range which is equal to or greater than 0 and is equal to or less than 1, and said evaluation value is an arbitrary value within a range which is equal to or greater than 0 and is equal to or less than 1.

9. A document retrieval system as claimed in claim 8, wherein said learning means has first means for modifying one or a plurality of said relationship values in said keyword connection table when said relevance value obtained by said calculation means is equal to 0, and second means for modifying one or a plurality of said relationship relation values in said keyword connection table when said relevance value obtained by said calculation means is not equal to 0.

10. A document retrieval system as claimed in claim 7, wherein said inputting means has a numerical value inputting part for inputting said evaluation value.

11. A document retrieval system as claimed in claim 10, wherein:

said learning means has (1) a difference calculating part for calculating a difference between said evaluation value from said numeral value inputting part and said relevance value from said ranking means, and (2) a modifying part for modifying one or a plurality of said relationship values in said keyword connection table based on the difference calculated by said difference calculating part; and

said one or a plurality of said relationship values are modified so that said relevance value increases when said evaluation value is greater than said relevance value and said relevance value decreases when said evaluation value is less than said relevance value.

12. A document retrieval system as claimed in claim 7, wherein said inputting means includes:

a rank designating part for designating an evaluation rank which indicates the degree of relevance required by the user for each of the documents output by said outputting means, the evaluation rank being indicated by a number; and

a converting part for converting said evaluation rank designated by said designating part into said evaluation value, the greater the evaluation rank indicating a correspondingly greater evaluation value is set.

13. A document retrieval system as claimed in claim 12, wherein said converting part converts said number of said rank into said evaluation value in accordance with the following formula, ##EQU35## where n is the number of said rank, N is the maximum number of said rank and f is said evaluation value.

14. A document retrieval system as claimed in claim 12, wherein:

said learning means has (1) a difference calculating part for calculating a difference between said evaluation value from said converting part and said relevance value from said ranking means; and

(2) a modifying part for modifying one or a plurality of said relationship values in said keyword connection table based on the difference calculated by said difference calculating part; and

said one or a plurality of said relationship values are modified so that said relevance value increases when said evaluation value is greater than said relevance value and said relevance value decreases when said evaluation value is less than said relevance value.

15. A document retrieval system as claimed in claim 12, wherein:

said designating part has a plurality of input elements, each of said input elements corresponding to one evaluation rank; and

when one of said plurality of input elements is designated by the user, a corresponding evaluation rank is supplied from said one of said plurality input elements to said converting part.

16. A document retrieval system as claimed in claim 15, wherein each of said plurality of input elements has an icon indicated on a display device.

3. A document retrieval system for retrieving documents from documents registered in a document database responsive to a retrieval condition designated by a user, said document retrieval system comprising:

a) query converting means for converting said retrieval condition designated by the user into a query having a predetermined normal form in which keywords and at least one type of logical operation out of logical operations AND, OR and NOT are connected;

b) bibliographical information indicating means for indicating at least information concerning which keywords correspond to each document of the documents registered in said document database;

c) a keyword connection table having relationship values, each of the relationship values representing a degree of relationship between keywords;

d) ranking means for ranking documents in accordance with relevance values, each relevance value indicating a degree to which a document fulfills the retrieval condition corresponding to the query, each of said relevance values being calculated for the document using the relationship values provided in said keyword connection table, wherein connected keywords are obtained with reference to the information indicated by said bibliographical indicating means as (1) keywords corresponding to the document and (2) keywords included in the query obtained by said query converting means;

e) outputting means for outputting, as a retrieval result, the documents ranked by said ranking means;

f) inputting means for inputting evaluation information indicating a degree to which each of the documents output by said outputting means is relevant to a document required by the user; and

g) learning means for modifying one or a plurality of said relationship values in said keyword connection table based on the evaluation information input by said inputting means;

wherein said query converting means includes second means for generating a second query in a disjunctive normal form, said second query in the disjunctive normal form having subqueries which are connected with each other only by logical OR operations, each of said subqueries being expressed by at least one keyword connected by at least one of logical AND and NOT operations;

wherein said ranking means includes;

1) first calculation means for calculating a sub-relevance value for each of said subqueries, said sub-relevance value indicating a degree to which each document fulfills each of said subqueries; and

2) second calculation means for calculating a relevance value for each of the documents using said sub-relevance value calculated for each of said subqueries by said first calculation means.

4. A document retrieval system as claimed in claim 3, wherein said second calculation means calculates said relevance value in accordance with the following formula, ##EQU34## where ri is said relevance value, ri(h) is said sub-relevance value calculated by said first calculation means and N is the number of said subqueries making up said query.