JP2000259637A - Information retrieving device, its retrieving method and recording medium storing information retrieval processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information retrieving device capable of simply and quickly retrieving object data without checking candidates diverged from a purpose by sorting, arranging and displaying a retrieved result on the basis of a previously defined concept keyword. SOLUTION: The information retrieving device is provided with an input means 1 such as a keyboard, a data base 2 for storing data, a retrieving means 3 for retrieving a data base on the basis of a retrieval request sentence applied from the input means 1, a concept knowledge recording means 4 for storing concept knowledge describing relation between data stored in the data base 2 and a concept keyword, and a sort/arrangement means 5 for sorting/arranging a retrieved result on the basis of the concept knowledge. Since a retrieved result is sorted/arranged by a concept keyword for reducing the deviation of data while referring to the concept knowledge and the sorted/arranged result of the retrieved result is displayed, object data can be simply and quickly retrieved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention provides a function for classifying and displaying information retrieved from a database using conceptual knowledge in a device having a built-in arithmetic device such as a computer, an OA device, a home electric appliance, etc. The present invention relates to an information search apparatus, an information search method, and a recording medium on which an information search processing program is recorded.

[0002]

2. Description of the Related Art When searching a database, if the number of hit data is large as a result of searching the database, an additional operation for narrowing down is required. In general, a keyword search is performed by adding a keyword. However, in order to effectively perform a search without excluding required data, a certain knowledge or technique is required. JP-A-6-124308
In Japanese Patent Application Publication No. 2000-209, search results are classified and arranged by using world knowledge, and support is provided so that an information search operation can be performed efficiently even if knowledge and skills related to information search are insufficient.

[0003]

SUMMARY OF THE INVENTION When classifying and organizing search results, if there is only one world knowledge model from a single viewpoint, the data in the database is biased or the structure of the world knowledge is sufficiently fragmented. If the search is not performed, there is a problem that the search results cannot be effectively sorted and arranged. The present invention provides a means for classifying search results more uniformly by having conceptual knowledge from multiple viewpoints, efficiently eliminating candidates that do not meet the purpose of the operator, and easily and quickly searching for the target data. The purpose is to:

[0004]

According to the present invention, there is provided an information retrieval apparatus comprising: an input means for inputting data and an instruction; a database storing various data; and a database comprising a search request sent from the input means. Search within
A search unit that generates a search result matching the search request from the database, and a concept that holds information of conceptual knowledge indicating a degree of a relationship between individual data stored in the predetermined database and a predetermined concept keyword. A knowledge storage unit; and a classification / arrangement unit for classifying and rearranging the search result by the search unit based on the concept knowledge in the concept knowledge storage unit.

[0005] Further, a word dictionary containing words and word vectors corresponding to the words, and a search result obtained from the search means, data of each of the search results and individual concept keywords in a set of concept keywords. And a concept knowledge generating means for generating the concept knowledge by determining the degree of the relationship with the word dictionary with reference to the word dictionary, thereby eliminating the need to generate the concept knowledge in advance for each data in the database. Therefore, it is particularly effective in a system where the database is not closed. Further, the concept knowledge storage unit holds concept knowledge information indicating the degree of relation to the concept keyword from a plurality of viewpoints, and the classification and sorting unit selects a concept keyword from a viewpoint that can classify data more uniformly. Thus, a result display with less data bias can be performed.

Further, the information retrieval method of the present invention retrieves a database storing various data using an input step of inputting data and instructions, and a request sentence given in the input step. A search step of generating a search result matching the request sentence from the database;
Categorizing and organizing the search results obtained by the search step based on information of conceptual knowledge indicating a degree of a relationship between the individual data stored in the predetermined database and a predetermined concept keyword. .

[0007] The present invention also provides an input step of inputting data and instructions to a computer, and searching a database storing various data using a request sentence given in the input step. A search step of generating a search result that matches a sentence from the database, and conceptual knowledge information indicating a degree of a relationship between a search result obtained by the search step and a predetermined concept keyword for individual data stored in the predetermined database. And a computer-readable recording medium that records a program for sequentially executing the steps of:

[0008]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 shows a first embodiment of the present invention. The search request sentence input from the input unit 1 such as a keyboard is passed to the search unit 3, and the search unit 3 searches the database 2 for data matching the search request sentence. The search results are sorted and arranged by the sorting and organizing means 5 with reference to the conceptual knowledge storage means 4, and the results are output by the output means 6 such as a monitor.

The following is a specific description of an example applied to a help system for a word processor. The input means 1 is a keyboard, and the output means 6 is a display. The database 2 stores help information, and holds information such as a help title, a description, and a search keyword. Various search methods such as simple keyword matching and analysis of the meaning of the input request sentence to search for the optimal help can be applied to the search means 3. The concept knowledge storage means 4 describes the relationship between individual data stored in the database 2 and concept keywords.
FIG. 3 is a diagram for explaining the structure of the concept knowledge storage unit 4. Now, four concept keywords, ie, “character / sentence”, “ruled line”, “cursor”, and “document / format” are defined as concepts for classifying help.

[0010] In the concept knowledge storage means 4, each data of the database 2 can be described as "1" when there is a relationship with these concept keywords, and "0" when there is no relationship with these concept keywords. If the data is "move the cursor up / down / left / right", "character / text", "ruled line", "document / format"
Is described as “0” because it is not related to “Cursor” and “1” because it is related to “Cursor”. If the arrangement of the four concepts is arranged in the order of (character / sentence, ruled line, cursor, document / format) and called a concept vector, in the case of the preceding data, the data is represented by the concept vector (0, 0, 1, 0). And the concept keywords can be described. Verification of each concept vector of data stored in the concept knowledge storage means 4 and each data stored in the database 2 can be easily achieved by assigning a common ID number to the data and the concept vector corresponding thereto. be able to.

The symbol 30 in FIG. 2 is an example of the top 20 candidates searched for in the search request sentence “I want to move”. The candidates 30 are arranged from left to right in the search order. These candidates express the relationship with the concept keywords 11 to 14 by connecting lines according to the concept vectors individually assigned. For example, “move the cursor up, down, left, and right” is connected only to the conceptual keyword “cursor” 13 because the concept vector is (0, 0, 1, 0). The sorting and organizing means 5 sorts and organizes the candidates 30 for each conceptual keyword based on this information and displays them on the output means 6 such as a display. 4 and 5 show examples of the display result. In this example, the search results are classified and displayed by tags provided for each conceptual keyword.

FIG. 4 shows a state in which, when a tag “character / sentence” is selected, only candidates associated with a conceptual keyword “character / sentence” are displayed as search results. Further, FIG. 5 shows a state in which when a tag “cursor” is selected, only candidates associated with the conceptual keyword “cursor” are displayed as search results.

FIG. 6 shows another embodiment of the information retrieval apparatus of the present invention. The input means 1, database 2, search means 3, concept knowledge storage means 4, classification and organization means 5, and output means 6 are shown in FIG.
Is the same as In FIG. 6, the concept keyword set 7,
A concept knowledge generating means 8 and a word dictionary 9 are added. No concept vectors have been generated in advance for the individual data in the database 2 in the concept knowledge storage means 4, and the concept knowledge generation means 8 receives a search result generated by the search means 3, and outputs the concept keywords in the concept keyword set 7. And dynamically generate conceptual knowledge. The word dictionary 9 describes words and their word vectors. The word vector has the same structure as the concept vector shown in FIG. 3, and expresses the meaning of the word in a distributed manner in association with the N feature words. The details of the word vector are described in JP-A-10-55372.

Hereinafter, a method of generating conceptual knowledge will be specifically described. FIG. 9 is a flowchart showing a method of generating conceptual knowledge from a search result. First, the processing is started (S
1). First, it is checked whether the number of search result candidates obtained from the search means 3 is not 0 (S2). If the result is 0, the process ends (S9). If it is not 0, the process proceeds to S3, and one candidate is extracted from the search result. Now, it is assumed that a candidate “move the cursor up and down” is extracted. Next, words are extracted from the candidates (S4). There are a method of extracting words by extracting a character string using a word dictionary and a method of using morphological analysis. Words such as “cursor”, “up / down”, “left / right”, “move” are extracted from the candidate “move the cursor up / down / left / right”. Next, the word vector of the extracted word is checked with reference to the word dictionary 9 (S
5). Subsequently, candidate sentence vectors are generated from the word vectors of these extracted words (S6). A method of generating a sentence vector is described in Japanese Patent Application Laid-Open No. H10-55372. Briefly, individual word vectors are added and the result is normalized. As a result, the sentence vector is one in which the meaning is dispersedly expressed in association with the N feature words, similarly to the word vector. When the sentence vector is determined, the degree of semantic similarity with each of the conceptual keywords defined in the conceptual keyword set 7 is calculated (S7). The semantic similarity can be represented by the magnitude of the inner product of the candidate sentence vector and the concept keyword word vector. It is determined that the greater the size, the greater the similarity. The detailed calculation method is described in JP-A-10-55372. Now, as a conceptual keyword "letter /
Assuming that sentences, "ruled lines", "cursors", and "documents / formats" have been defined, these word vectors are obtained with reference to the word dictionary 9, and the inner product of the sentence vectors obtained in S6 is obtained. The semantic similarity with each concept keyword is obtained. Subsequently, 1 is given to a concept vector whose semantic similarity exceeds a certain threshold, and 0 is given to other concept vectors, to generate a candidate concept vector (S8). The value of the threshold value is set by examining an appropriate value by which data in the database 2 can be appropriately classified by the concept vector based on experiments. As a result of this processing, the concept vector of “moving the cursor up, down, left, and right” is obtained as (0, 0, 1, 0) as shown in FIG. When the process is completed up to S8, the process returns to S2 to continue the process. As a result of this series of processing, the same number of corresponding concept vectors as the search result is generated. These are passed as conceptual knowledge to the classification and arrangement means 5 and used for classification and arrangement of the search results. In this embodiment, the concept knowledge of the concept knowledge storage means 4 does not need to be generated in advance for each piece of data in the database 2, which is particularly effective in a system where the database is not closed.

FIG. 8 is a diagram for explaining the conceptual knowledge of the third embodiment. The configuration of the system is common to the first embodiment shown in FIG. 1 and the second embodiment shown in FIG.
The difference is that the concept knowledge of the concept knowledge storage means 4 is configured from a plurality of viewpoints. In FIG. 8, two viewpoints, “object” and “function”, are defined. Concept keywords are defined for each viewpoint, and “character / text”, “ruled line”, “cursor”, “document / format” are used for the “target” viewpoint.
Are defined as conceptual keywords, and from the viewpoint of “function”, conceptual keywords such as “move”, “input”, “delete”, and “copy” are defined. If the concept knowledge storage means 4 is represented by a structure of (object (character / sentence, ruled line, cursor, document / format), function (move, input, delete, copy)), the data of "moving the cursor up / down / left / right" The concept vector looks like ((0,0,1,0), (1,0,0,0)).

The symbol 30 in FIG. 7 is an example of the top 20 candidates searched for in the search request sentence “I want to move”.
The candidates 30 are arranged from the left in the order of the search order.
These candidates express the relationships with the concept keywords 11 to 14 and 21 to 24 by connecting lines according to the concept vectors individually assigned. In the case of the data of "moving the cursor up, down, left, and right", the concept keyword "cursor" 13 and the concept keyword "move" 21 are connected according to the concept vector. In the sorting and organizing means 5, candidate 3
0 is classified and arranged by concept keyword classification from a more effective viewpoint. In the case of FIG. 7, since the candidate 30 is a hit as a result of the search request sentence “I want to move”,
Many related to the concept keyword 21 of “movement” from the viewpoint 20 of “function”. On the other hand, "target"
In the concept keywords 11 to 14 of the viewpoint 10, related candidates are more uniformly distributed than those of the viewpoint 20 of “function”. As a result, the classification means 5 performs classification based on the viewpoint 10 of “object”, that is, the concept keyword “character /
The candidates 30 are sorted and arranged based on “sentence” 11, “ruled line” 12, “cursor” 13, and “document / format” 14, and the result by the output unit 6 is displayed. FIG. 4 and FIG. 5 are examples of the result display. Note that the present invention is not limited to the above embodiment.

Each process of the information retrieval device described above is realized by a program for executing a target function. This program is stored in a computer-readable recording medium. In the present invention, the recording medium is a program medium (not shown), and the program and data in the program medium are read by being mounted on a reading device (not shown).

The program medium is a recording medium that is configured to be separable from the main body, such as a tape system such as a magnetic tape or a cassette tape, a magnetic disk such as a floppy disk or a hard disk, or a CD-ROM / MO / MD / D.
Disk system of optical disk such as VD, card system such as IC card / optical card, or mask ROM, EPROM,
It may be a medium that fixedly carries a program including a semiconductor memory such as an EEPROM or a flash ROM.

[0019]

The search results of the database can be classified and displayed according to the concept, so that it is easy to find necessary results from many search results. In addition, by having a plurality of viewpoints, even if data or search results are biased, search results can be optimally classified and displayed by selecting an appropriate viewpoint. As a result, the user interface is greatly improved as compared with the conventional search result display based on hit rank.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a first embodiment of an information search device according to the present invention.

FIG. 2 is a configuration diagram showing a relationship between data and concept keywords.

FIG. 3 is a diagram showing a structure of conceptual knowledge.

FIG. 4 is a diagram showing an example of display of search results sorted and arranged;

FIG. 5 is a diagram showing an example of display of search results sorted and arranged;

FIG. 6 is a block diagram illustrating a configuration of a second embodiment of the information search device of the present invention.

FIG. 7 is a diagram illustrating a relationship between data and a concept keyword when a plurality of viewpoints are provided.

FIG. 8 is a diagram showing a structure of conceptual knowledge when a plurality of viewpoints are provided.

FIG. 9 is a flowchart for generating conceptual knowledge.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input means 2 Database 3 Search means 4 Concept knowledge storage means 5 Classification organization means 6 Output means 7 Concept keyword set 8 Concept knowledge generation means 9 Word dictionary 10, 20, Perspective keywords 11-14, 21-24, Concept keywords S1-S9 Processing steps

Claims (5)

[Claims] An input means for inputting data and instructions,
A database in which various data are stored, a search unit that searches the database with a search request sentence provided by the input unit, and generates a search result that matches the search request from the database; A concept knowledge storage unit for holding information of concept knowledge indicating a degree of a relationship between the individual data stored in the database and a predetermined concept keyword, and a search result by the search unit for storing the concept knowledge in the concept knowledge storage unit. An information retrieval apparatus, comprising: a classification / arrangement unit configured to classify / arrange based on information. 2. A word dictionary containing words and word vectors corresponding to the words, and individual concept keywords in a set of individual data of the search results and concept keywords from a search result obtained from the search means. 2. The information retrieval apparatus according to claim 1, further comprising: conceptual knowledge generating means for generating the conceptual knowledge by determining the degree of the relationship with the word dictionary while referring to the word dictionary. 3. The concept knowledge storage means holds concept knowledge information indicating the degree of relation to the concept keyword from a plurality of viewpoints, and the classification and organizing means stores the concept keyword from the viewpoint of more uniformly classifying data. The information retrieval apparatus according to claim 1, wherein the information retrieval apparatus is selected. 4. An inputting step of inputting data and instructions, and searching a database storing various data using a search request sentence given in the inputting step.
A search step of generating a search result matching the search request sentence from the database, and a concept indicating a degree of a relationship between a search result obtained by the search step and a predetermined concept keyword with respect to individual data stored in the predetermined database. An information retrieval method, comprising: a classification and arrangement step of classifying and organizing based on knowledge information. 5. An input step of inputting data and instructions to a computer, and searching a database storing various data using a search request sentence given in the input step, and matching the search request sentence. A search step of generating a search result obtained from the database, and a search result obtained by the search step based on information of concept knowledge indicating a degree of relationship between individual data stored in the predetermined database and a predetermined concept keyword. A computer-readable recording medium on which is recorded a program for sequentially executing a sorting and sorting step for sorting and sorting.