JP2002108936A - Information retrieving device, control method therefor and computer readable storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information retrieving device, a control method therefor and a computer readable storage medium, with which an operator can select a desired similar retrieving method out of similar retrieving methods of plural kinds at the viewpoint of similarity. SOLUTION: The operator inputs a retrieval sentence 303 for retrieving desired multimedia information in order to retrieve that information and selects any one of radio buttons 305-307 capable of selecting the viewpoint in the case of similar retrieval. When a retrieve button 308 is operated, according to the viewpoint selected by the operator, the information retrieving device retrieves multimedia information matched to the retrieval sentence 303 or having similarity based on the relevant viewpoint out of a database storing a plurality of multimedia information beforehand.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval apparatus which manages a plurality of multimedia information, retrieves desired multimedia information from the managed multimedia information, a control method thereof, and a computer-readable storage. Regarding the medium.

[0002]

2. Description of the Related Art Conventionally, in the field of information retrieval,
For example, there has been proposed an information search device for searching for so-called multimedia information in which a document or character describing information content is added to information such as an image. As such a device, for example, desired information is searched for. In response to the operator's input of a word or sentence to be represented, a word or sentence that matches the input word or sentence is searched (acquired) from a large number of pieces of multimedia information stored in a database in advance. 2. Description of the Related Art There is known an information search device to which a so-called full text search method is applied.

In recent years, a word or sentence that matches a word or sentence input by an operator is not only searched for, but also a word or sentence determined to be similar to the input word or sentence is detected. An information search device that applies a so-called similar concept search method for searching for held multimedia information has also been proposed.

[0004] Furthermore, an information retrieval apparatus to which a visual similarity search for retrieving a document or image in a format having similarity such as a form by focusing on the layout and structure of the document instead of the content of the document is applied. Has also been proposed.

[0005]

However, in the above-mentioned conventional information retrieval apparatus, similar information retrieval capable of retrieving multimedia information having similar contents is not limited to color information of an image, style of a document, layout or the like. Since there are a plurality of types of similar information search methods such as a similar information search that can search for visually similar multimedia information, an operator can select a multimedia information that he or she wants to obtain. It is difficult to determine from what viewpoint similar information is searched for information, and operability is a problem.

Accordingly, the present invention provides an information retrieval apparatus which allows an operator to select a desired similarity retrieval technique from a plurality of types of similarity retrieval techniques from the viewpoint of similarity, a control method thereof, and a computer-readable storage medium. For the purpose of providing.

[0007]

In order to achieve the above object, an information retrieval apparatus according to the present invention has the following configuration.

That is, an information search apparatus for searching for desired multimedia information from a plurality of pieces of multimedia information stored in a database in advance, wherein a search request for searching for desired multimedia information can be input. Input means, for the search request input from the input means,
A first similarity information search unit that searches for similar multimedia information from the database based on a first similarity viewpoint, and a second viewpoint different from the first similarity for the search request. A second similarity information search unit that searches for similar multimedia information from the database, and a similar viewpoint selection unit that can select a similarity viewpoint when searching for the desired multimedia information,
An information search control unit that controls the first or second similar information search unit to perform an information search in accordance with the similarity viewpoint selected by the similar viewpoint selection unit.

[0009] In a preferred embodiment, the first similar information search means measures similarity of the search request with character information expressed in a language attached to the plurality of multimedia information. The visual similarity may be measured based on the layout information attached to the multimedia information, or the similarity of the metadata attached to the plurality of pieces of multimedia information may be measured.

[0010] The first similar information search means may include, as the metadata, a title of the information, a creator of the information, a manager of the information, Affiliation information of the creator of the information, who can use the information, the date and time of creation of the information, a summary sentence of the description of the information, a description language of the information, a description format of the information, a usage fee of the information, Preferably, at least one of the data sizes of the information is used.

[0011] In order to achieve the above object, a control method of an information retrieval apparatus according to the present invention has the following configuration.

That is, the present invention relates to a method for controlling an information retrieval apparatus for retrieving desired multimedia information from a plurality of pieces of multimedia information stored in advance in a database, wherein a retrieval request for retrieving the desired multimedia information is provided. An input step to be input, and a similar viewpoint selection step in which a similar viewpoint to be used as a reference when searching for the desired multimedia information is selected from among a plurality of types of similar viewpoints. And a similar information search step of searching the database for similar multimedia information based on the similarity viewpoint selected in the similar viewpoint selection step.

[0013] In a preferred embodiment, in the similarity viewpoint selecting step, the similarity information searching step includes:-an information search process based on a viewpoint of similarity with linguistically expressed character information accompanying the plurality of pieces of multimedia information; The information search processing is performed based on visual similarity based on the layout information attached to the plurality of pieces of multimedia information; or It is preferable that a similarity viewpoint can be selected so that an information search process is performed based on a similarity viewpoint of metadata attached to information.

Further, the present invention is characterized by a computer-readable storage medium storing a program code for realizing the above-mentioned information retrieval apparatus and its control method by a computer.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the drawings as an embodiment applied to an information retrieval apparatus for retrieving document information.

<Apparatus Configuration> FIG. 1 is a block diagram showing an apparatus configuration of an information retrieval apparatus according to the present embodiment.

In FIG. 1, reference numeral 11 denotes a microprocessor (CPU) which performs an operation for information retrieval, logical judgment, etc., and comprises an address bus AB, a control bus CB,
Each component connected to these buses is controlled via a data bus DB.

The address bus AB transfers an address signal designating a component to be controlled by the CPU 11. The control bus CB transfers a control signal of each component to be controlled by the CPU 11. The data bus DB performs data transfer between the constituent devices.

Reference numeral 12 denotes a read-only fixed memory (RO)
M), and stores a control program code such as a processing program executed in the present embodiment.

Reference numeral 13 denotes a writable random access memory (RAM) composed of 16 bits per word, which is used for temporarily storing various data to be transferred to each component and various received data. Also, the RAM 13
Is a search term holding unit 202, a search viewpoint holding unit 203, a developed word holding unit 207, a search concept feature 1 holding unit 210, a search concept feature 2 holding unit 225, and a search result holding unit 213, which will be described later with reference to FIG. Remember.

Reference numeral 14 denotes an external memory (DISK) as a large-capacity storage device, which is a search word expansion dictionary 206, a concept dictionary 208, and a stored document concept feature 1 described later with reference to FIG.
Holding unit 209, stored document conceptual feature 2 holding unit 218, stored document form holding unit 219, stored document attribute holding unit 22
0, stored document 217, and word index 226 are stored.

Here, in the present embodiment, the stored document 2
Reference numeral 17 denotes a database storing a plurality of documents (document data files) to be searched in a document search process (FIG. 9) described later. The registered document 224 is a document (document data file) to be processed in a document registration process (FIG. 12) described later in order to prepare various information necessary for the document search process. This is a document arbitrarily selected from the stored documents 217 when the document registration process is executed.

Further, in the external memory 14, a search request input processing unit 201, a similar search processing determination unit 204, a search word expansion processing unit 205, a conceptual feature 1 extraction processing unit 209, a related word Extraction processing unit 211, similar concept feature search processing unit 214, similar form search processing unit 215, similar attribute search processing unit 216, concept feature 2 extraction processing unit 22
1. A program code for executing each processing unit of the document form determination processing unit 222 and the document attribute extraction processing unit 223 is stored.

As storage media for storing data and programs stored in the external memory 14, ROM, floppy (registered trademark) disk, CD-ROM, DVD
-ROMs, memory cards, magneto-optical disks and the like can be used.

Reference numeral 15 denotes a keyboard (KB), such as an alphabet key, a hiragana key, a katakana key, a character symbol input key such as a period, a search key for instructing a search, and a cursor movement key for instructing a cursor movement. Function keys.

Reference numeral 16 denotes a display video memory (VRAM).
And temporarily stores the data pattern to be displayed.

17 is a CRT controller (CRTC)
And plays a role of displaying the drawing contents stored in the display video memory VRAM 16 on the display device 18.

Reference numeral 18 denotes a display device (CRT) such as a cathode ray tube, a liquid crystal panel, etc., and a display pattern of a dot configuration and a display of a cursor on the display device CRT are controlled by a CRT controller 17.

Reference numeral 19 denotes a network controller (NI
C), and the information retrieval device according to the present embodiment is Ethe
It plays the role of connecting to networks such as rnet.

The information retrieval apparatus having the above-described configuration operates according to various inputs by the operator via the keyboard KB15 and various inputs via the network supplied from the network controller 19. That is, when an input from the keyboard KB15 and / or an input from the network controller 19 are supplied to the information search device (FIG. 1), first, an interrupt signal is sent to the microprocessor CPU11, and the microprocessor C
The PU 11 reads out various control signals stored in the external memory DISK 14 and according to the control signals,
Various controls are performed.

A storage medium in which a program related to a search operation described in the present embodiment is stored in advance is shown in FIG.
, And the computer of the device reads and executes the program code stored in the storage medium, thereby achieving the search operation in the present embodiment.

<Functional Configuration> Next, a functional configuration of the information retrieval apparatus according to the present embodiment will be described with reference to FIG.

FIG. 2 is a block diagram showing a functional configuration of the information search device according to the present embodiment.

In FIG. 2, reference numeral 201 denotes a search request input processing unit in which requirements (a search sentence or a search word, a similar search viewpoint, etc.) regarding a desired search target input by the operator via the keyboard 15 are set. It is. 202
Is a search word holding unit that stores the search words input to the search request input processing unit 201. A search viewpoint holding unit 203 stores the similar search viewpoint input to the search request input processing unit 201.

A similarity search processing determination unit 204 determines the type of similarity search process to be executed according to the similarity search viewpoint stored in the search viewpoint holding unit 203.

Reference numeral 206 denotes a search word expansion dictionary in which a word serving as a heading and a related word to be expanded with respect to the word are stored together with the degree of relevance between the two. 2
Reference numeral 05 denotes a search word expansion processing unit that expands search words or search sentences stored in the search word holding unit 202 into related words that can be expanded by referring to the search word expansion dictionary 206. Reference numeral 207 denotes a developed word holding unit that stores related words developed by the search word development processing unit 205 for the search word to be searched.

Reference numeral 208 denotes a concept dictionary that stores a concept in which the semantic characteristics of a word serving as a heading are described. 209
Is a concept feature 1 extraction processing unit that extracts the concept feature of the search request sentence or the registered document as the concept feature 1 by referring to the expanded word holding unit 207, the concept dictionary 208, and the registered document 224.

Reference numeral 210 denotes a search concept feature 1 holding unit that holds the concept feature 1 for the search term 202 extracted by the concept feature 1 extraction processing unit 209. 225
Is a search concept feature 2 holding unit that holds the concept feature 2 for the search word 202 extracted by the concept feature 2 extraction processing unit 221.

Reference numeral 211 denotes a related word extraction processing unit that extracts a set of highly relevant words from each document stored in the stored document 217 and stores the extracted word set in the search word expansion dictionary 206. .

Reference numeral 221 denotes a concept feature 2 extraction processing unit for extracting the concept feature 2 from the registered document 224. 218
Is a storage document concept feature 2 holding unit that stores the concept feature 2 extracted by the concept feature 2 extraction processing unit 221 in association with the registered document 224. Reference numeral 212 denotes a storage document concept feature 1 holding unit that holds the concept feature 1 corresponding to the registered document 224 extracted by the concept feature 1 extraction processing unit 209.

Reference numeral 214 denotes a stored document conceptual feature 1 holding unit 21
2. A similar concept feature search processing unit that executes a similar concept search process with reference to the search concept feature 1 holding unit 210 and the stored document concept feature 2 holding unit 218.

Reference numeral 222 denotes a document form determination processing unit that determines the document form of the registered document 224. 219 is
A stored document form holding unit that stores the document form determined by the document form determination processing unit 222 in association with the registered document 224. A similar form search processing unit 215 searches for a similar form document by referring to the stored document form holding unit 219.

Reference numeral 223 denotes a document attribute extraction processing unit for extracting the document attributes of the registered document 224. Reference numeral 220 denotes a document attribute extracted by the document attribute extraction processing unit 223,
24 is a stored document attribute storage unit that stores the stored document attribute in association with 24.
A similar attribute search processing unit 216 searches for a similar attribute document by referring to the stored document attribute holding unit 220.

Reference numeral 213 denotes a search result holding unit that holds the processing results of each of the similar concept feature search processing unit 214, the similar form search processing unit 215, and the similar attribute search processing unit 216.

Next, in the search request input processing unit 201, a display example of an operation panel when the operator instructs a search request sentence or a search request word of a similar search and a viewpoint from which to execute a similar search will be described. This will be described with reference to FIG.

FIG. 3 is a diagram showing a display example of the operation panel in the case of instructing a search request sentence or a search request word and a similar search viewpoint in this embodiment.

In FIG. 3, reference numeral 301 denotes a display window for performing an input operation to make a search request. Reference numeral 302 denotes a search sentence input area for inputting a sentence or a word as a target of a search request. Reference numeral 303 denotes an example of a search request sentence being input by the operator, and in the figure, "mobile device market trend" is input. An input cursor 304 indicates an input position in the search sentence input area.

Reference numerals 305, 306, and 307 denote radio buttons that allow the operator to specify a similarity determination viewpoint in similar document search. More specifically, the document content similarity 305 is used in terms of the meaning of the document. Specify when instructing to search for similar documents. Document form similar 306
Is specified when a search for a document having a similar format and format is performed. The document attribute similarity 307 is specified when instructing a search for documents having similar document bibliographic attribute information such as title, creator, creation date and time, manager, classification, and keyword. In the present embodiment, when the operator does not give an explicit instruction using these radio buttons, the document content similarity 305 is implicitly selected.

Reference numeral 308 denotes a search execution button for designating execution of the search processing. When the search execution button 308 is pressed, the specified search processing is executed. Reference numeral 309 denotes a cancel button for designating termination or cancellation of the search processing. When the cancel button 309 is pressed, the search processing is immediately terminated, and the display window 301 is closed and terminated.

Reference numeral 310 denotes a search result display area for displaying the result of the search processing performed by pressing the search button 308. In the figure, nothing is displayed since the search processing is not performed.

Next, by the processing of the search request input processing unit 201, the operator instructs the search request sentence or the search request word of the similar search on the screen of FIG. The case where the search process is executed in accordance with these instruction items will be described with reference to FIG.

FIG. 4 is a diagram showing a display example of a search result in the present embodiment, a search request sentence or a search request word or a similar request document designation when a similar search is to be subsequently performed, and an input of a similar search viewpoint. .

In FIG. 4, reference numeral 401 denotes a search result display area corresponding to 310 in FIG. Reference numeral 402 denotes a rank display area indicating the rank of a search result. In the present embodiment, the search results are ranked in an order similar to the search request, and displayed in the rank order. In the display example of FIG.
Search results from the 25th to 30th ranks are displayed.

At 403, the title of the retrieved document is displayed, and at 404, the file name of the retrieved document is displayed. In 405, the content of the retrieved document is displayed to the extent that it can be understood. That is, in the document content display 405, an abstract sentence given in advance as a bibliographic attribute of the document, an abstract automatically summarized from the document, a rough sentence extracted from a part of the document as an outline, and the like are displayed. .

Reference numeral 406 designates a display position in order to display another part which is not displayed while partially displaying the search result in the display area 401 when the display result cannot be completely displayed in the search result display area 401. This is the elevator bar used in the same type of window display device.

The display state illustrated in FIG. 4 is a search in which the document content similarity search designated by the document content similarity radio button 305 is performed on “the market trend of the mobile device” indicated in the search sentence 303. The result is displayed. At this time, when the operator continues to perform a similarity search, he or she inputs a search sentence in the search request sentence input area 302 as described above, or displays a display line of a document for which a similar search is desired from the search result display area 401. Is selected, and by dragging and dropping the selected display line to the search request sentence input area 302, the document content of the selected document may be copied (not shown) to the search request sentence input area 302.

<Search Word Expansion Dictionary 203> Next, the configuration of the search word expansion dictionary 203 will be described with reference to FIG.

FIG. 5 is a diagram exemplifying a configuration of a search word expansion dictionary in the present embodiment.

As shown in FIG. 5, the search word expansion dictionary 203 includes an ID 501 indicating identification information of expansion information, an expansion headword 502, and an expansion word 50 related to the expansion headword 502.
3. The degree of association 50 between the expanded headword 502 and the expanded word 503
4. Each item of the FLG 505 indicating the type of the expanded word is stored in advance in a state where the items are associated with each other.

That is, the ID 501 is the search expansion dictionary 203
Is an identification number uniquely assigned to each record stored in the. The expansion entry word 502 stores a word to be an expansion entry word. The expansion word 503 is an expandable word such as an association word or a related word related to the expansion headword 502. The degree of relevance 504 indicates the degree of relevance (relevance) of the expansion word 503 corresponding to the expansion headword 502 by a numerical value from 0 to 9, and indicates that the higher the stored numerical value is, the higher the relevance is. Show. The FLG 505 stores the expansion word type (that is, the expansion word 503 corresponding to the expansion headword 502).
Is identification information indicating whether the word is an initially registered word prepared in advance as a standard or a related word extracted from a stored document), and “S” is a standard expansion registered in the initial registration. The word “A” indicates a related word extracted from the stored document.

<Concept Dictionary 208> Next, the concept dictionary 208
Will be described with reference to FIG.

FIG. 6 is a diagram showing the configuration of the concept dictionary in this embodiment.

The concept dictionary 208 is a representation of a concept of a word as a multidimensional vector having elements of universal semantic weights as shown in FIG. 6. As shown in FIG. The weight (weight value) 603 of each element of the semantic vector is stored in a state corresponding to the subscript 602 representing each element of the semantic vector expressed in 256 dimensions. The weight 603 of the semantic vector takes a real number between 0 and 1, and is normalized and stored so that the size of the semantic vector becomes 1.

Here, a universal semantic element which is an element of a multi-dimensional vector constituting the concept dictionary 208 is a label representing a single semantic concept, for example, “this, that, that, which”. Such as "concept of instruction", "class, grade, grade, rank, ranking,
Words such as Inferiority, Inferiority, and Honor are included in the concept of "grade", and words such as "change, transformation, innovation, and rise" are included in the concept of change, "cooperation, greeting, unity, Such as "concept of communion", which includes words such as "handshake, friendship, diplomatic relations, and companionship", and "concept of life", which includes words such as "animal, mammal, penguin, dog, human, goldfish". In this way, each independent universal semantic that does not depend on a specific word is used. In the present embodiment, as shown in FIG. 6, 256 types of semantics are used to form a 256-dimensional concept expression vector.

<Word Index 226> Next, the structure of the word index 226 will be described with reference to FIGS.

FIG. 7 is a diagram showing the structure of a word index in this embodiment.

The word index 226 is the registered document 22
4 is a table that stores the frequency of appearance of all words that appear in the document. In FIG. 7, reference numeral 701, which is the first column information of the table, is a document ID for uniquely identifying a registered document. The second column information to the n-th column information 702 of the table are suffixes representing each word shown in FIG. The last row 703 of the table stores the sum of the number of occurrences of each word. In FIG. 7, the document ID is 00
The document 146 indicates that the word “market” indicated by the subscript 1256 appears 12 times in the document.

Next, in the word index 226,
The configuration of a correspondence table storing correspondence between words and subscripts in the word index table will be described with reference to FIG.

FIG. 8 is a diagram showing the structure of a correspondence table between word indexes and words in the present embodiment. In the correspondence table, a unique word index 802 associated with a word 801 is associated. For example, the word index of the word “market” is 1256.

<Document Similarity Search Process> Next, a document similarity search process executed in this embodiment will be described with reference to FIG.

FIG. 9 is a flowchart showing a document similarity search process executed by the information search device (FIG. 2) in the present embodiment.

In the figure, in step S1001,
A search request input process is performed by a processing module that performs the operation of the search request input processing unit 201 in FIG. still,
Details of the search request input processing will be described later.

In step S1002, step S10
In the process 01, the similarity search processing type is determined according to the similarity search viewpoint stored in the search viewpoint holding unit 203 in FIG. That is, the document concept similarity search process (step S100)
3), document form similarity search processing (step S100)
4), branch to any of document attribute similarity search processing (step S1005).

In step S1003, a document concept similarity search process for searching for a document having a similar concept represented by the document contents is performed according to the search request input in the search request input process (step S1001). The details of the document concept similarity search process will be described later.

In step S1004, a document form similarity search process is performed to search for a document whose document format is similar to the document ID input in the search request input process (step S1001). This document form similarity search process employs a currently common similarity search method, and a description in this embodiment is omitted.

In step S1005, a document attribute similarity search process is performed to search for documents in which the document ID input in the search request input process (step S1001) and the bibliographic attribute information of the document are similar. The document attribute similarity search process employs a currently common similarity search method for searching metadata of a document, and a description in this embodiment is omitted.

In step S1006, step S10
03, the search results searched in any of the steps S1004 and S1005 are retrieved from the search result holding unit 213 in FIG. 2 and displayed.
It should be noted that the processing for displaying the search result employs a general method at present, and the description in this embodiment is omitted.

(Search Request Input Processing) Next, step S1
Details of the search request input process 001 will be described with reference to FIG.

FIG. 10 is a flowchart showing details of the search request input process in the embodiment of the present invention.

In the figure, in step S2001,
As a result of the operator's instruction operation, the search sentence to be the target of the search request is, as described above with reference to FIG.
2 is input at the position indicated by the input cursor 304. Here, the search sentence may be a single word or a sentence composed of a plurality of words. Input of search sentence is KB15
Is stored in the search word holding unit 202 on the RAM 13.

Next, in step S2002, it is determined whether or not an input has been made to the search sentence input area 302. If there is no search sentence input to the search sentence input area 302, any of the search result documents displayed in the search result display area 401 If it is detected that is dragged to the search sentence input area 302,
The process branches to step S2003, and if a search sentence input is detected, the process branches to step S2006.

In step S2003, the document file name of the document dragged to the search text input area 302 is obtained from the information displayed in the file name 404 (see FIG. 4), and the document corresponding to the file name is read. Open for

Next, in step S2004, step S
The content of the document opened in 2003 is read, and in step S2005, the read document content is stored as a new search request sentence in the search word holding unit 202, and the content is displayed in the search request sentence input area 302. .

In step S 2006, the viewpoint from which similar document search is to be performed is determined based on the document content similar button 305, document form similar button 306, and document attribute similar button 307.
Is determined based on which one has been checked, the selected similarity search viewpoint is stored in the search viewpoint holding unit 203, and the process ends.

(Document Concept Similarity Search Process) Next, the details of the document concept similarity search process in step S1003 will be described with reference to FIG.
1 will be described.

FIG. 11 is a flowchart showing details of the document concept similarity search process in the present embodiment.

In step S3001, the search sentence stored in search word holding unit 202 in the above-described search request input process (FIG. 10) (the sentence input by the operator to search for a desired document, or the re-search Is extracted, and the character string constituting the retrieved search sentence is divided into words (search words). To divide the search sentence into words (search words), a technique known as morphological analysis processing is applied.

Next, in step S3002, step S
With respect to all the search words obtained in 3001, expansion word data that matches the expansion entry word 502 of the search expansion dictionary 206 is searched. If there is a headword that matches the search word, the expansion word 503 is set to the corresponding relevance degree 5
04 and the search word are all stored in the expanded word holding unit 207.

In step S3003, step S30
In step 02, the expansion word stored in the expansion word holding unit 207 is extracted, and a search is made to see if there is any word that matches the entry word 601 (FIG. 6) in the concept dictionary 208. If there is a concept dictionary entry word 601 that matches the expanded word, the corresponding concept vector data is extracted from the concept dictionary 208.

In step S3004, the obtained value is obtained by multiplying the component value of the component of the acquired concept vector data by the value of the degree of association 504 stored in the expanded word holding unit 207 in correspondence with the expanded word. By adding the product to the search request concept feature 1, search request concept feature 1 data is constructed as first similarity evaluation scale information for evaluating similarity. In the search request concept feature 1, all the dimension elements of the vector are initialized to 0 in advance.

In step S3005, expanded word holding unit 2
It is determined whether or not all expansion words stored in 07 have been processed. When all expansion words have been processed, the search request concept feature 1 vector data is normalized so that the sum of squares of each element becomes 1. After the conversion, it is stored in the search concept feature 1 holding unit 210, and the process branches to step S3006. If there is an unprocessed expansion word, the process loops to step S3003.

Next, in step S3006, it is determined whether or not the search request concept feature 1 vector data has been constructed, based on whether or not the search request concept feature 1 vector data is a zero vector. If so, the process branches to step S3010,
On the other hand, if the search request concept feature 1 vector data has been constructed, the flow branches to step S3007.

At step S3007, at step S30
In the determination of step 06, it was determined that the search request concept feature 1 vector data was not zero and that the search request concept feature 1 vector data could be constructed, so the search request concept feature 1 vector data stored in advance in step S5007 of the document registration process (FIG. 13) described later. , The corresponding stored document concept feature 1 vector data is acquired from the stored document concept feature 1 holding unit 212.

In step S 3008, the concept similarity between the acquired stored document concept feature 1 vector data and the search request concept feature 1 vector data stored in the search request concept feature 1 holding unit 210 is calculated. In this step,
The concept similarity can be calculated from the cosine measure of the two vector data. The calculated concept similarity is
The data is temporarily stored in a buffer memory (not shown) in association with the document ID of the stored document.

In step S3009, it is determined whether or not the processing in step S3008 has been completed for all the stored documents.
7 and if all stored documents have been processed, the flow branches to step S3016.

On the other hand, if it is determined in step S3006 that the search request concept feature 1 vector data remains zero and that the search request concept feature 1 vector data cannot be constructed, step S3006 described below is performed.
Through the processing of each step from step 3010 to step S3015, search request concept feature 2 vector data is constructed, and a similar concept search of the document content is performed by calculating the concept similarity with the concept feature 2 vector data of the stored document.

That is, in step S3010, a developed word is extracted from the developed word holding unit 207.

Next, in step S3011, with reference to the word index 226, vector data having the word index as a dimension element is constructed as search request concept feature 2 data. The search request concept feature 2 data is second similarity evaluation scale information for evaluating similarity, which is an evaluation scale different from the search request concept feature 1 described above. That is, word ID 8 is obtained from the word index table.
For words that 02 can be obtained as an index,
The occurrence frequency of the word is given as an element value with the index as a subscript, and 0 is given as an element value with the index of the word that did not appear as a subscript.

In step S3012, expanded word holding unit 2
In step S3010, it is determined whether or not all expanded words stored in 07 have been processed. If there are unprocessed expanded words, the process loops to step S3010. Are normalized so that the sum of the squares of the respective element values becomes 1 and the process branches to a search requesting step S3013.

In step S3013, the stored document concept feature 2 vector data stored in advance in step S5007 of the document registration process (FIG. 13) described later is extracted from the stored document concept feature 2 holding unit 218. Feature 2 of search request concept constructed by
Of the dimensional elements of vector data, stored document concept feature 2
If any of the dimension elements of the vector data does not exist, the component element value corresponding to the word index is set to 0 and added as an element of the stored document concept feature 2 vector data.

In step S3014, step S30
13, the stored document conceptual feature 2 vector data acquired in
The concept similarity with the search request concept feature 2 vector data calculated in step S3011 is calculated. The calculation of the concept similarity in this step can be obtained from the cosine measure of the two vector data. The calculated concept similarity is temporarily stored in a buffer memory (not shown) in association with the document ID of the stored document.

In step S3016, all search results of similar concepts calculated in step S3008 or S3014 and stored in a buffer memory (not shown) are sorted in descending order of concept similarity, and the sorted results are sorted. Stored in the search result holding unit 213,
finish.

<Document Registration Processing> Next, the document registration processing will be described with reference to FIG. In this embodiment,
The document registration process is executed regularly or irregularly at a maintenance position separately from the document search by the above-described document search process in order to ensure efficient information search by the information search device.

FIG. 12 is a flowchart showing a document registration process executed in this embodiment.

In the figure, in step S4001,
A document content concept registration process for extracting and registering concept features from the content of the registered document is performed. The details of the document content concept registration process will be described later.

Next, in step S4002, a process of extracting a search related word from the stored document 217 including the registered document and registering it in the search word expansion dictionary is performed. The details of the search-related word extraction processing will be described later.

Next, in step S4003, the concept dictionary 2
Concept feature 1 is estimated for a search word that is not registered as a headline word 601 in 08 (FIG. 6), and a search word concept extraction process for performing additional registration is performed. The details of the search term concept extraction processing will be described later.

In step S4004, a document form registration process for registering document form information is performed based on document format information such as the style, appearance, and format of the registered document. This process performs a known process in the same type of information search device.

In step S4005, document attribute registration processing for registering the metadata, the author, the administrator, the creation date and time, the creation department, the creation purpose, and other metadata accompanying the document as document attributes of the registered document is performed. This process performs a known process in the same type of information search device.

In step S4006, the target registration document is registered in the registration document 224 in a state where the various data obtained in the steps S4001 to S4005 are associated with each other.

(Document Content Concept Registration Process) Next, the details of the document content concept registration process in step S4001 will be described with reference to FIG.
3 will be described.

FIG. 13 is a flowchart showing details of the document concept registration process in this embodiment.

In FIG. 17, in step S 5001,
The words forming the character string of the registered document 224 (target registered document) to be registered are extracted. For the extraction of words, a method generally used for morphological analysis is applied.

Next, in step S5002, step S500
Each word extracted in 5001 is registered in the word index 226. That is, if the word matches the word 801 in the word index table, the corresponding word ID is acquired, and the appearance frequency in the target registered document is stored in the column information of the table using the acquired word ID as an index. On the other hand, for a word that does not match the word 801 in the word index table, a new unique word ID is assigned to the word in order to add the word to the word index table, and the assigned word ID is stored in a column of the table. In addition to the information column, 0 is stored as a frequency for a registered stored document, and an appearance frequency is stored for a target registered document.

Next, in step S5003, document concept feature 2 vector data is constructed with reference to the word index 226. That is, the line data of the target registered document of the word index 226 is extracted, and the appearance frequency is given as a component element for each word index.

In step S5004, step S50
01 is searched in the concept dictionary 208 for a headword 601 that matches the word extracted in 01, and as a result of the search,
If there is a matching headword, the corresponding concept feature 1 vector data is extracted.

In step S5005, in step S50
The product calculated by multiplying the conceptual feature 1 vector data extracted in step 04 by a weight corresponding to the frequency is calculated in step S500.
3 is added to the stored document conceptual feature 2 vector data. Here, the weight multiplied by the frequency is given according to the product of the word frequency in the document (TF value) and the inverse document frequency (IDF).

In step S5006, it is determined whether or not processing has been completed for all words in the registered document. If there is an unprocessed word, the processing loops to step S5001, and if processing has been completed for all words, processing branches to step S5007. .

In step S5007, the document concept feature 1 vector data is normalized so that the sum of squares of the vector elements becomes 1, and the normalized data is converted to the data shown in FIG.
Stored document conceptual feature 1 used in step S3008
As vector data, stored document conceptual feature 1 holding unit 21
Register in 2. Also, the document concept feature 2 vector data is similarly normalized such that the sum of squares of the elements becomes 1, and the lengths thereof are made uniform. It is registered as the concept feature 2 vector data in the stored document concept feature 2 holding unit 218, and the processing ends.

(Search Related Word Extraction Process) Next, step S
Details of the search-related word extraction processing 4002 will be described with reference to FIG.

FIG. 14 is a flowchart showing details of the search-related word extraction processing in this embodiment.

In the figure, in step S6001,
The words constituting the character string of the registered document 224 (target registered document) are extracted. For the extraction of words, a method generally used for morphological analysis is applied.

In step S6002, for each extracted word, the degree of correlation with other words appearing in the registered document 224 is obtained, and a co-occurring word having a high relevance is obtained. As a method of calculating the degree of correlation, the word index table 22
Referring to FIG. 6, words appearing specifically in the target registered document that co-occur with the word of interest, although the frequency of appearance in other documents is low, are referred to as word frequency (TF value) and inverse document frequency (TF value). I
DF value) is determined by ranking from the product with the largest product.

In step S6003, step S60
02 is stored in the search word expansion dictionary 206 as the expansion word 5
03, it is searched whether it has already been registered. As a result of the search, if it has been registered, the process loops to step S6002 to acquire the next co-occurrence word.
Processing branches to step S6004.

In step S6004, step S60
In 03, the co-occurrence word determined not to be registered in the search word development dictionary 206 is additionally registered in the search word expansion dictionary 206 as a development word corresponding to the word. At this time, the co-occurrence correlation calculated in step S6002 is proportionally divided into a value of the relevance 504,
FLG 505 is registered in association with flag information “A” indicating an additional expansion word.

In step S6005, it is determined whether or not all co-occurring words have been processed.
Loop to 6002.

In step S6006, it is determined whether all the words of the registered document 224 have been processed. If there is an unprocessed word, the process loops to step S6001. If all the words have been processed, the process ends.

(Search Word Concept Extraction Processing) Next, step S
Details of the search term concept extraction processing 4003 will be described with reference to FIG.

FIG. 15 is a flowchart showing details of the search term concept extraction processing according to the embodiment of the present invention.

In step S7001, the concept dictionary 208
An unknown concept word that is not registered in the headword 601 is acquired. Here, to determine whether or not the word is an unknown concept word, a word that exists in the word 801 of the word index 226 and does not exist in the headword 601 of the concept dictionary 208 is determined to be an unknown concept word.

In step S7002, a related word of the extracted unknown concept word is obtained by searching the search word expansion dictionary 206, and the obtained related word, an expansion word having the relevant word as an expansion headword 502, and both Is stored in the expanded word holding unit 207.

In the step S7003, the expanded word holding unit 2
The related word stored in 07 is acquired, and the concept dictionary 208 is searched according to the acquired related word, thereby acquiring the concept feature 1 vector data corresponding to the related word. At this time, if the related word is not registered as a headword in the concept dictionary 208, a zero vector in which all the elements are 0 is set as the concept feature 1 vector data.

In step S7004, step S70
In the concept feature 1 vector data of the related word obtained in 03,
The product obtained by multiplying the degree of association stored in the expanded word holding unit 207 is added to the concept feature 1 vector data of the unknown concept word.
The concept feature 1 vector data of the unknown concept word is initialized in advance to a zero vector in which all element values are 0.

In the step S7006, the expanded word holding unit 2
It is determined whether all the related words stored in 07 have been processed. If there is any unprocessed related word, the process loops to step S7003. If all the related words have been processed, the process branches to step S7007.

In step S7007, the unknown word concept feature 1 vector is normalized so that the sum of squares of each element value becomes 1, and then registered in the concept dictionary 208.

In step S7008, it is determined whether or not all unknown concept words have been processed. If there is an unprocessed unknown concept word, the process loops to step S7001. If all unknown concept words have been processed, the process ends. .

According to the above-described embodiment, the operation instruction for similarity information search is combined into one operation screen as illustrated in FIG. 3 (FIG. 4), and which similar information search is to be performed. In this configuration, the operator can explicitly and easily specify the concept from the viewpoint of the similarity, so that an appropriate concept similarity search can be performed, and the convenience is improved.

[0138]

[Other Embodiments] In the above embodiment, description was made using document information as a search target. However, multimedia information such as image information, moving image information, and program content description information accompanied by content description metadata other than document information is described. Also, the present invention can be applied by extracting the information feature amount from the text information in which the content is described, and calculating the similarity measure of the information feature amount.

In the above-described embodiment, the stored document 217 to be searched and the stored document conceptual feature 1 holding unit 2
12, storage document concept feature 2 storage unit 218, storage document form storage unit 219, storage document attribute storage unit 220, word index 226, concept dictionary 208, search word expansion dictionary 2
06 has been described as being arranged in the DISK 14 constituting a single device, but it is also possible to distribute these components to different devices and perform processing on a network via the NIC 19. .

Even if the present invention is applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), a device including one device (for example, a copying machine, a facsimile machine, etc.) ) May be applied.

Further, an object of the present invention is to provide a storage medium storing a program code of software for realizing the functions of the above-described embodiments to a system or an apparatus, and to provide a computer (or CPU or MPU) of the system or apparatus.
PU) reads out and executes the program code stored in the storage medium.

In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, hard disk, magneto-optical disk, CD-ROM, CD-R, CD
-RW, DVD-ROM, DVD-RAM, DVD-R
W, DVD + RW, magnetic tape, nonvolatile memory card, ROM and the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) Performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, based on the instruction of the program code, The CPU provided in the function expansion board or function expansion unit performs part or all of the actual processing,
The case where the function of the above-described embodiment is realized by the processing is also included.

[0146]

As described above, according to the present invention,
An information search apparatus, a control method thereof, and a computer-readable storage medium capable of allowing an operator to select a desired similarity search technique from the viewpoint of similarity among a plurality of types of similarity search techniques are provided.

[Brief description of the drawings]

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

FIG. 2 is a block diagram illustrating a functional configuration of the information search device according to the embodiment.

FIG. 3 is a diagram illustrating a display example of an operation panel when a search request sentence or a search request word and a similar search viewpoint are instructed in the present embodiment.

FIG. 4 is a diagram showing a display example of a search request sentence and a search result corresponding to a similar search viewpoint in the embodiment.

FIG. 5 is a diagram exemplifying a configuration of a search word expansion dictionary in the embodiment.

FIG. 6 is a diagram showing a configuration of a concept dictionary in the present embodiment.

FIG. 7 is a diagram illustrating a configuration of a word index according to the present embodiment.

FIG. 8 is a diagram showing the correspondence between word IDs and words in a word index according to the embodiment.

FIG. 9 is a flowchart illustrating a similar document search process executed in the embodiment.

FIG. 10 is a flowchart illustrating details of a search request input process executed in the embodiment.

FIG. 11 is a flowchart illustrating details of a document concept similarity search process executed in the present embodiment.

FIG. 12 is a flowchart illustrating a document registration process executed in the embodiment.

FIG. 13 is a flowchart illustrating details of a document content concept registration process executed in the present embodiment.

FIG. 14 is a flowchart illustrating details of a search-related word extraction process executed in the present embodiment.

FIG. 15 is a flowchart illustrating details of a search term concept extraction process executed in the present embodiment.

[Explanation of symbols]

 11: CPU, 12: ROM, 13: RAM, 14: DISK, 15: KB, 16: VRAM, 17: CRTC, 18: CRT, 19: NIC,

Claims (12)

[Claims] 1. An information search device for searching for desired multimedia information from a plurality of multimedia information stored in a database in advance, wherein a search request for searching for the desired multimedia information can be input. Input means; first similarity information search means for searching for similar multimedia information from the database based on a first similarity viewpoint with respect to the search request input from the input means; A second similarity information searching means for searching the database for similar multimedia information based on a second viewpoint different from the first similarity; and a similarity when searching for the desired multimedia information. Similar viewpoint selection means capable of selecting a viewpoint of gender; Other information retrieval apparatus characterized by comprising, an information search control means for controlling so that the information retrieval is performed by the second similarity information retrieval means. 2. The method according to claim 1, wherein the first similar information search unit measures the similarity of the search request with linguistically expressed character information accompanying the plurality of multimedia information. Information retrieval device. 3. The method according to claim 1, wherein the first similar information search unit measures visual similarity of the search request based on layout information attached to the plurality of pieces of multimedia information. Information retrieval device. 4. The information search apparatus according to claim 1, wherein the first similar information search unit measures similarity of metadata attached to the plurality of pieces of multimedia information with respect to the search request. . 5. The first similar information search means, as the metadata, a title of the information, a creator of the information, a manager of the information, Affiliation information of the creator of the information, who can use the information, the date and time of creation of the information, a summary sentence of the description of the information, a description language of the information, a description format of the information, a usage fee of the information, The information retrieval apparatus according to claim 4, wherein at least one of the data sizes of the information is used. 6. A control method of an information search device for searching for desired multimedia information from a plurality of pieces of multimedia information stored in a database in advance, wherein a search request for searching for the desired multimedia information is provided. An input step to be input, and a similarity point selection step in which a similarity point to be used as a reference when searching for the desired multimedia information is selected from a plurality of similarity points of view. A similar information search step of searching the database for similar multimedia information based on the similarity viewpoint selected in the similar viewpoint selection step for the input search request. Control method of search device. 7. In the similarity viewpoint selection step, an information search process is performed in the similarity information search step based on a viewpoint of similarity with linguistically expressed character information accompanying the plurality of pieces of multimedia information. 7. The method according to claim 6, wherein a similarity viewpoint can be selected. 8. In the similarity viewpoint selection step, in the similarity information search step, an information search process based on a viewpoint of visual similarity is executed based on layout information attached to the plurality of pieces of multimedia information. 7. The method according to claim 6, wherein the similarity viewpoint can be selected. 9. The similarity viewpoint selecting step includes: executing a similarity search so that an information search process based on a similarity viewpoint of metadata attached to the plurality of pieces of multimedia information is performed in the similarity information search step. 7. The method according to claim 6, wherein a viewpoint can be selected. 10. In the similar information search step, a title of the information, a creator of the information, a manager of the information, and affiliation information of a creator of the information, which are attached to the plurality of pieces of multimedia information. , The user who can use the information, the creation date and time of the information, a summary sentence of the description of the information, the description language of the information, the description format of the information, the usage fee of the information, and the data size of the information. 10. The control method according to claim 9, wherein any one is used as the metadata. 11. A computer-readable storage medium storing a program code for operating a computer as the information search device according to claim 1. Description: 12. A computer-readable storage medium storing a program code capable of realizing a control method of the information search device according to claim 6 by a computer.