JP6390139B2 - Document search device, document search method, program, and document search system - Google Patents

Description

  The present invention relates to a document search technique for searching for contents such as documents and books suitable for a user.

  In recent years, a matching method for recommending a document according to a user's preference to a user who has selected a document or a book has been researched and developed. As existing main matching methods, a method of recommending documents with similar contents (document base), a method of recommending a document desired by another user who has selected the same document (cooperation base), and a user attribute There is a method (rule base) for recommending a corresponding document.

  On the other hand, Patent Document 1 describes a technique for extracting an unexpected sentence from a document.

JP 2011-95905 A

  However, when existing matching methods are used, known documents such as the same series are often presented, or multiple documents with too high similarity are often presented, and the recommendation result has no unexpectedness or diversity. There was a problem.

  According to Patent Document 1, an unexpectedness score is given to sentences in a search target document based on the number of appearances of each word appearing in a document selected by the user (hereinafter referred to as a search target document). Although there is a description, there is no description that the document is searched based on the unexpectedness score.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to search for a document having unexpectedness and diversity in addition to similarity to a document specified by a user. It is to provide a search device and the like.

In order to solve the above-described problem, a first invention is a document search device that searches for a recommended document that is related to a search target document to be searched, and manages a feature word for each document of the recommended document, First storage means for storing, second storage means for managing and storing feature words for each category of recommended documents, and documents of the target document and recommended documents with reference to the first storage means A document similarity calculation unit that calculates a document similarity that is a similarity in units, and a classification similarity that is a similarity in a classification unit between the target document and the recommended document with reference to the second storage unit Classification similarity calculation means for calculating the difference, and a search means for taking the difference between the document similarity and the classification similarity and extracting the recommended document with reference to the calculated difference value. This is a document retrieval device.
According to the first invention, it is possible to search for a recommended document having unexpectedness and diversity with respect to a search target document as a starting point. Further, since the recommended document can be extracted based on the relationship between the similarity of the document and the similarity of the classification, it is possible to search for a recommended document having unexpectedness and diversity.
The “classification” is a collection of recommended documents of the same type and corresponds to a category in the embodiment.

In addition, it is desirable to extract a case where the difference value exceeds a preset threshold value.
This makes it possible to extract a recommended document having a high document similarity but a low classification similarity, and therefore, it is possible to search for recommended documents having unexpectedness and diversity.

Further, it is desirable that the search means obtains a difference by multiplying the classification similarity by a predetermined coefficient.
Thereby, the similarity of classification can be adjusted by a predetermined coefficient, and a recommended document whose classification is somewhat similar can be searched, or a recommended document whose classification is not similar can be searched.

It is preferable that the search unit searches the recommended document using the document similarity and / or the classification similarity satisfying a preset threshold.
As a result, dissimilar documents can be excluded or classifications with high similarity can be excluded, so that it is possible to search for recommended documents with unexpectedness and diversity.

According to a second aspect of the present invention, there is provided a computer-based document search method for searching for a recommended document having a relation to a search target document to be searched, wherein the control unit of the computer manages a feature word for each document of the recommended document. A document similarity calculating step for calculating a document similarity, which is a similarity in document units, between the target document and the recommended document with reference to the first storage means, and a feature word for each category of the recommended document A classification similarity calculating step of calculating a classification similarity that is a similarity in a classification unit between the target document and the recommended document, with reference to the second storage means that manages and stores the document similarity, The document search method includes a search step of taking a difference from the classification similarity and extracting the recommended document by referring to the calculated difference value .
According to the second invention, it is possible to search for a recommended document having unexpectedness and diversity with respect to a search target document as a starting point. Further, since the recommended document can be extracted based on the relationship between the similarity of the document and the similarity of the classification, it is possible to search for a recommended document having unexpectedness and diversity.

A third invention is a program for causing a computer to function as a document search device that searches for a recommended document that is related to a search target document to be searched. The computer is characterized for each document of the recommended document. First storage means for managing and storing words, second storage means for managing and storing feature words for each category of recommended documents, and referring to the first storage means, the target document and the recommended document A document similarity calculating means for calculating a document similarity that is a similarity in document units with the second storage means, and a classification that is a similarity in a classification unit between the target document and the recommended document. A program for functioning as classification similarity calculation means for calculating similarity, a search means for taking the difference between the document similarity and the classification similarity, and extracting the recommended document with reference to the calculated difference value It is.
According to the third invention, it is possible to search for a recommended document having unexpectedness and diversity with respect to a search target document as a starting point. Further, since the recommended document can be extracted based on the relationship between the similarity of the document and the similarity of the classification, it is possible to search for a recommended document having unexpectedness and diversity.

A fourth invention is a document search system in which a user terminal and a document search device that searches for a recommended document that is related to a search target document to be searched are connected via a network, wherein the user terminal is An input receiving unit that receives an input of the search target document, a transmission unit that transmits the search target document to the document search device, and a display unit that receives and displays a search result from the document search device, The document search apparatus includes a first storage unit that manages and stores a feature word for each document of a recommended document, a second storage unit that manages and stores a feature word for each category of a recommended document, and the user Document similarity calculation for calculating a document similarity, which is a similarity in document units between the target document and the recommended document with reference to the receiving unit that receives the search target document from the terminal and the first storage unit Means and With reference to the second storage unit, a classification similarity calculation means for calculating a classification similarity is similarity classification units of the target document and recommendations document, and the classification similarity between the document similarity A document search system comprising: search means for extracting the recommended document with reference to the calculated difference value; and transmission means for transmitting the extracted recommended document to the user terminal. It is.

  With the document search device of the present invention, it is possible to search for documents with unexpectedness and diversity in addition to similarity to documents specified by the user.

1 is a diagram illustrating an example of a system configuration of a document search system according to the present embodiment. The figure explaining the outline | summary of the document search service which concerns on this embodiment 1 is a block diagram showing a hardware configuration example of a document search device (user terminal) according to the present embodiment. The figure which shows an example of the category information memorize | stored in a category information database The figure which shows an example of the document information memorize | stored in a document information database The figure which shows an example of the information temporarily hold | maintained at a document search device Flow chart showing the flow of document recommendation processing Illustration explaining the document vector Figure showing an example of a document recommendation screen Flow chart showing the flow of category similarity calculation processing Flow chart showing the flow of document similarity calculation processing Flow chart showing the flow of unexpectedness score calculation processing The flowchart which shows the flow of the document recommendation process of the document search apparatus which is not connected to the network

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
First, the configuration of the present embodiment will be described with reference to FIGS.

  FIG. 1 is a diagram illustrating an example of a system configuration of a document search system 100 according to the present embodiment. As shown in FIG. 1, the document search system 100 is configured by a document search apparatus 1 and one or a plurality of user terminals 2 (2a, 2b) connected to each other via a network 3.

  In the present embodiment, a document registered in the document information database 6 is searched for a document (search target document) selected by a user using the user terminal 2 based on similarity to the search target document, and the like. An example in which the document search system 100 according to the present invention is used in the document recommendation service presented to the user terminal 2 will be described. FIG. 2 is a diagram for explaining the outline of the document recommendation service.

  In the present invention, a document is an electronic book, magazine, article, paper, other document, article content published on the Internet, or the like.

  As shown in FIG. 2, in the document recommendation method a (conventional document recommendation service), a document having a high similarity to the search target document 10 (document A) is searched from a database and recommended. In this case, the similarity between the search target document 10 and the recommendation destination document 11 (document B, document C, document D) is high, and the similarity between the search target document 10 and the category 12 (category X) to which the recommendation destination document belongs. As a result, the result is high. Therefore, in the document recommendation method a, the recommendation destination document 11 and the search target document 10 have the same classification and similar contents, so that a document with little unexpectedness is presented to the user.

  On the other hand, in the document recommendation method b, a document having a low similarity to the search target document 10 (document A) is searched from the database and recommended. In this case, the similarity between the search target document 10 and the recommendation destination document 11 (document E) is low, and the similarity between the search target document 10 and the category 12 to which the recommendation destination document belongs (category Y) is also low. It becomes. Therefore, in the document recommendation method b, the recommendation destination document 11 and the search target document 10 are classified so far that their contents are not similar, so that an unrelated document is presented to the user.

  The document recommendation service according to the present invention is shown in the document recommendation method c, and has high similarity with the search target document 10 (document A) and low similarity with the category 12 (category Z) to which the recommended document belongs. A document (document F) is retrieved from the database and recommended. As a result, it is possible to present to the user a document with a different classification but similar contents, that is, a surprising document.

Returning to the description of FIG.
The document search device 1 is a server device of a site that provides a document recommendation service. The storage unit 22 of the document search device 1 has a category information database (DB) 5 that manages category information to which the document belongs, and a document that manages document information. An information database (DB) 6 and the like are held. In addition, the document search apparatus 1 holds user history information that manages the identification information of the user terminal 2 that uses the document recommendation service and the history of purchasing (or browsing, etc.) documents (for example, books) in association with each other. Also good. Details will be described later.

  The user terminal 2 is a computer used by a user who uses the document recommendation service, and displays a document recommendation screen 80 (see FIG. 9) transmitted from the document search device 1. The user terminal 2 may be a mobile terminal, a mobile terminal, or the like instead of a general-purpose computer.

  FIG. 3 is a hardware configuration diagram of a computer that realizes the document search apparatus 1 (user terminal 2) according to the embodiment of the present invention. As shown in FIG. 3, the computer includes a control unit 21, a storage unit 22, a media input / output unit 23, a communication control unit 24, an input unit 25, a display unit 26, a peripheral device I / F unit 27, etc. 28 is connected and configured.

The control unit 21 includes a CPU (Central Processing Unit) and a ROM (Read Only).
Memory), RAM (Random Access Memory) and the like.
The CPU calls and executes a program stored in the storage unit 22, ROM, storage medium or the like in the work memory area on the RAM, drives and controls each device connected via the bus 28, and the document search device 1 ( The processing to be described later performed by the user terminal 2) is realized. The ROM is a non-volatile memory and permanently holds a computer boot program, a program such as BIOS, data, and the like. The RAM is a volatile memory, and temporarily stores a loaded program, data, and the like, and includes a work area used by the control unit 21 to perform each process.

  The storage unit 22 is an HDD (Hard Disk Drive) or the like, and stores a program executed by the control unit 21, data necessary for program execution, an OS (Operating System), and the like. These program codes are read by the control unit 21 as necessary, transferred to the RAM, and read and executed by the CPU.

The media input / output unit 23 is a media input / output device such as a CD drive, a DVD drive, an MO drive, and a floppy (registered trademark) disk drive, and inputs / outputs data such as images.
The communication control unit 24 includes a communication control device, a communication port, and the like, and is a communication interface that mediates communication between the computer and the network 3, and performs communication control with other devices via the network 3. The network 3 may be wired or wireless.

The input unit 25 performs data input and includes, for example, a keyboard, a pointing device such as a mouse, and an input device such as a numeric keypad. The input data is output to the control unit 21.
The display unit 26 includes, for example, a display device such as a CRT monitor or a liquid crystal panel, and a logic circuit (video adapter or the like) for executing display processing in cooperation with the display device, and is input by the control of the control unit 21. The displayed information is displayed on the display device.
The input unit 25 and the display unit 26 may be, for example, a display with a touch panel in which those functions are integrated.

The peripheral device I / F unit (interface) 27 is a port for connecting a peripheral device to the computer, and the computer transmits and receives data to and from the peripheral device via the peripheral device I / F unit 27. The peripheral device I / F unit 27 is configured by USB, IEEE 1394, RS-232C, or the like, and usually includes a plurality of peripheral devices I / F. The connection form with the peripheral device may be wired or wireless.
The bus 28 is a path that mediates transmission / reception of control signals, data signals, and the like between the devices.

  FIG. 4 is a diagram illustrating an example of category information 50 handled by the category information database 5. As shown in FIG. 4, the category information 50 includes information such as “category code 51”, “category name 52”, “category description 53”, “feature word 54”, and the like of each category. The category code 51 is an identifier for uniquely identifying a category, such as a Japanese decimal code number for classifying books, a genre registered for classifying article documents, and the like. The category name 52 is a heading representing the content of the category. The category explanation sentence 53 is a sentence explaining the category, and the feature word 54 is a word (keyword) representing the content of the category and its importance (described later).

  It should be noted that the category information 50 does not hold the feature word 54 but based on the category description 53 and the feature word 65 of each document belonging to the category, the feature word 54 of the category information 50 is executed every time document recommendation processing described later is executed. May be extracted.

  FIG. 5 is a diagram showing an example of document information 60 handled by the document information database 6. As shown in FIG. 5, the document information includes information such as “document code 61”, “document name 62”, “category code 63”, “document description 64”, and “feature word 65” of each document. It is a waste. The document code 61 is an identifier that uniquely identifies a document. The document name 62 is a document title. The category code 63 indicates the category to which the document belongs, and is associated with the category code 51 of the category information database 5. Documents registered in the document information database 6 are classified and registered in advance into one or more categories.

  The document description 64 is an introductory sentence representing the contents of the document, and the feature word 65 is a word (keyword) representing the contents of the document and its importance (described later). The document information 60 may extract the feature word 65 every time a document recommendation process (to be described later) is executed from the document description 64 or the document body without holding the feature word 65.

  FIG. 6 is a diagram illustrating an example of information temporarily stored in the document search apparatus 1 in the document recommendation process. As shown in FIG. 6A, the RAM or the like of the control unit 21 of the document search apparatus 1 stores “category code 71” and “search target” calculated in the category similarity calculation process (S106 in FIG. 7). The category similarity 72 ”with the document is linked and stored as category similarity information 56. The category code 71 is associated with the category code 51 of the category information database 5.

  As shown in FIG. 6B, the RAM or the like of the control unit 21 of the document search apparatus 1 stores “document code 73” and “search target” calculated in the document similarity calculation process (S107 in FIG. 7). The document similarity 74 with the document and the “unexpectedness score 75 with the search target document” calculated in the unexpectedness score calculation process (S108 in FIG. 7) are linked and stored as the document similarity information 57. . The document code 73 is associated with the document code 61 of the document information database 6.

[Document recommendation process]
Next, document recommendation processing executed by the document search device 1 and the user terminal 2 will be described with reference to FIGS.

FIG. 7 is a flowchart illustrating an example of the document recommendation process.
When the user terminal 2 accesses the document search apparatus 1, the control unit 21 of the document search apparatus 1 transmits a document selection reception screen (not shown) to the user terminal 2. On the document selection reception screen, documents registered in advance in the document information DB 6 are displayed as a list or a thumbnail, and the control unit 21 of the user terminal 2 receives the selection of the document by a user operation (step S101).

  The control unit 21 of the user terminal 2 transmits the document code 61 of the selected document (search target document) to the document search device 1 (step S102). The control unit 21 of the document search apparatus 1 receives the document code 61 of the search target document (step S103).

  The control unit 21 of the document search apparatus 1 extracts the keyword of the feature word 65 corresponding to the received document code 61 and the number of appearances of the keyword from the document information DB 6 (step S104), and generates a document vector of the search target document. (Step S105).

Specifically, the control unit 21 of the document search apparatus 1 performs morphological analysis on each document document (document name 62, document description 64, document content) using general-purpose morpheme analysis software, Extract keywords. A document vector is generated by calculating the TF of the extracted keyword.
Note that TF (Term-Frequency) is the frequency (number of appearances) that a keyword appears in a document.

  The control unit 21 of the document search apparatus 1 calculates the category similarity indicating the similarity with the document group belonging to the same category for each category registered in the category information DB 5 (step S106), and the search target document and other documents are calculated. Is calculated for each document registered in the document information DB 6 (step S107), and an unexpected score between the search target document and another document is calculated for each document (step S108). Details of each process executed in steps S106 to S108 will be described later.

  Based on the unexpectedness score 75 (stored in the document similarity information 57) with the search target document calculated in step S108, the control unit 21 of the document search apparatus 1 selects a document with a high unexpectedness score from the document information DB 6. (Step S109) and transmit them to the corresponding user terminal 2 as the recommendation destination document 11 (step S110).

  The control unit 21 of the user terminal 2 receives the data transmitted from the document search device 1 (step S111), displays the document recommendation screen 80 on the display unit 26 (step S112), and ends the process.

  FIG. 9 shows an example of the document recommendation screen 80. As shown in FIG. 9, the document recommendation screen 80 includes a search target document 10 (document A) selected by the user and a recommendation destination document 11 (document B, document G, document K) searched by the document recommendation process. List display or thumbnail display.

As described above, when receiving the selection of the search target document from the user terminal 2 by the document recommendation process, the document search device 1 generates a document vector representing the characteristics of the search target document. Subsequently, the similarity between the document group belonging to each category registered in the category information DB 5 and the search target document is calculated, and the similarity between each document registered in the document information DB 6 and the search target document is calculated. The document search apparatus 1 belongs to a category having a low similarity to the document selected by the user based on the above, but searches for a document having similar content between the documents, and uses the searched document as a recommended destination document. To present.
Thereby, the document search device 1 can recommend a document with unexpectedness and diversity to the document selected by the user.

The document recommendation process is not limited to the above-described one, and can be changed without departing from the spirit of the document recommendation process.
For example, in step S101 to step S103, the user terminal 2 is made to select a search target document from documents registered in advance in the document information DB 6 of the document search apparatus 1, but in the user terminal 2, as the search target document, An arbitrary character string input is accepted from a general search window, and the document search device 1 receives it, or the user terminal 2 uploads (sends) the created document to the document search device 1 to search the document. The apparatus 1 may receive it. The document search apparatus 1 extracts keywords from the received search target document, calculates importance, and generates a document vector of the search target document.

[Category similarity calculation processing]
Next, an example of the category similarity calculation process executed in step S106 of the document recommendation process will be described with reference to FIG. 10 and FIG. 8 as appropriate. In the category similarity calculation process, the control unit 21 of the document search device 1 calculates the similarity between the search target document and each comparison target category.

  First, in step S201, the control unit 21 of the document search apparatus 1 generates a document vector from the search target document.

  A method for generating the document vector of the search target document in step S201 will be specifically described with reference to FIG. Note that the document A in FIG. 8A corresponds to the search target document in the present embodiment.

The control unit 21 of the document search apparatus 1 generates a document vector of the document A from the “keyword” of the document A and the “number of appearances” of the keyword.
For example, in FIG. 8A, “Keyword” is “Moon, Earth, Satellite, Osaka, Nagoya, Baseball”, and “Number of appearances” is a number described in the table. At this time, the document vector Va of the document A is Va = “2, 1, 1, 0, 0, 1”.

  Next, in step S202, the control unit 22 of the document search apparatus 1 generates a document vector of each comparison target category.

  A method for generating a document vector of a category to be compared will be specifically described with reference to FIG. As shown in FIG. 8B, it is assumed that category X has sentence B and document C.

  Similarly to the generation of the document vector Vа of the document A, the control unit 21 of the document search apparatus 1 determines the document vector Vb and the document C of the document B from the “keyword” of the document B and the document C and the “number of appearances” of the keyword. Document vector Vc is generated.

  For example, in FIG. 8, “keyword” is “month, earth, satellite, Osaka, Nagoya, baseball”, and “number of appearances” is a number described in the table. At this time, the document vector Vb of the document B is Vb = “1, 1, 2, 0, 0, 2”, and the document vector Vc of the document C is Vc = “1, 0, 1, 0, 0, 0”. Become.

  Finally, the document vector of category X is the sum of the document vector Vb of document B and the document vector Vc of document C, that is, the document vector Vx of category X is Vx = “2, 1, 3, 0, 0, 2 ".

  The document vectors of the above categories are generated by the control unit 22 for all comparison target categories.

  In step S203, the control unit 21 of the document search device 1 calculates the similarity (category similarity) between the document vector Va generated in step S201 and the document vector of each comparison target category generated in step S202. To do.

  For example, when calculating the similarity with the document vector Vx of category X, the control unit 21 of the document search apparatus 1 calculates the vector Va and the vector Vx from the document vector Va of the document A and the document vector Vx of the category X. Is calculated (category similarity). As the category similarity, for example, cosine similarity (Va · Vx) / (| Va || Vx |) can be used. For example, in the case of FIG. 8, the category similarity is about 0.89.

  In step S <b> 203, the control unit 21 of the document search device 1 associates the calculated category similarity 72 with the category code 71 of the comparison target category and associates it with the category similarity information 56 to store the storage unit 22 of the document search device 1. Or it memorize | stores in RAM etc. of the control part 21. FIG.

  When the similarity between the search target document and all comparison target categories is calculated in step S203, the control unit 21 of the document search device 1 ends the process.

  As described above, through the category similarity calculation process, the document search apparatus 1 obtains the characteristics of each category from the document vector (document vector of the category) representing the characteristics of the document group belonging to the same category, and the similarity with the document selected by the user. Is digitized. As a result, when searching for recommended content, it is possible to extract content belonging to a category having low similarity to the document selected by the user.

  In the present embodiment, TF is used to generate a document vector. However, the present invention is not limited to this, and the document vector may be generated by TF-IDF (Term Frequency-Inverse Document Frequency).

[Document similarity calculation processing]
Next, an example of the document similarity calculation process executed in step S107 of the document recommendation process will be described with reference to FIG.

FIG. 11 is a flowchart illustrating an example of a document similarity calculation process.
The control unit 21 of the document search apparatus 1 inputs a comparison target document that is one document registered in the document information DB 6 (step S401).

  The control unit 21 of the document search apparatus 1 extracts the feature word 65 of the comparison target document from the document information DB 6 (step S402), and generates a document vector of the comparison target document (step S403). The control unit 21 of the document search apparatus 1 calculates the document similarity from the document vector of the search target document generated in S105 and the document vector of the comparison target document generated in Step S403 (Step S404). As the document similarity, the above-described cosine similarity may be used.

  The control unit 21 of the document search device 1 associates the calculated document similarity 74 with the document code 73 of the comparison target document to generate the document similarity information 57 as the storage unit 22 of the document search device 1 or the RAM of the control unit 21. (Step S405), and it is determined whether all the documents registered in the document information DB 6 have been input (step S406).

  If the input has not been completed (NO in step S406), the control unit 21 of the document search device 1 returns to step S401. If it has been input (YES in step S406), the control unit 21 of the document search device 1 ends the process.

  As described above, according to the document similarity calculation process, the document search apparatus 1 obtains the feature of each document from the document vector expressing the feature of each document, and quantifies the similarity with the document selected by the user. This makes it possible to extract a document similar to the document selected by the user when searching for recommended documents.

[Surprising score calculation process]
Next, an example of the unexpectedness score calculation process executed in step S108 of the document recommendation process will be described with reference to FIG.

FIG. 12 is a flowchart illustrating an example of the unexpectedness score calculation process.
The control unit 21 of the document search apparatus 1 inputs a comparison target document that is one document registered in the document information DB 6 (step S501).

  The control unit 21 of the document search apparatus 1 searches for the category code to which the comparison target document belongs, and extracts the category similarity 72 with the search target document from the category similarity information 56 based on the category code (step S502).

  The control unit 21 of the document search apparatus 1 extracts the document similarity 74 with the search target document from the document similarity information 57 based on the document code of the comparison target document (step S503).

  The control unit 21 of the document search apparatus 1 calculates the unexpectedness score 75 of the search target document with respect to the comparison target document using the following formula (1) (step S504).

  The weighting coefficient α in Expression (1) is set so that the unexpectedness score of a document belonging to a category having a low similarity to the search target document among documents having a high similarity to the search target document is high. The weighting coefficient α is a value set as appropriate based on the number of categories and the number of documents belonging to the category.

  The control unit 21 of the document search apparatus 1 stores the calculated unexpectedness score in the corresponding document similarity information 57 (step S505), and determines whether all the documents registered in the document information DB 6 have been input. (Step S506).

  If the input has not been completed (NO in step S506), the control unit 21 of the document search apparatus 1 returns to step S501. If it has been input (YES in step S506), the control unit 21 of the document search apparatus 1 ends the process.

  As described above, according to the unexpectedness score calculation process, the document search device 1 searches for a document belonging to a category having a low similarity to the search target document among documents having a high similarity to the search target document. Can be calculated. Thereby, when searching for a recommended document, a document having a high unexpectedness score with respect to a document selected by the user can be extracted from the document information DB 6 and a document with unexpectedness and diversity can be presented to the user. .

  Note that the document that is selected by the user in step S101 of the document recommendation process may be a document that is not registered in the document information DB 6. For example, it may be an article document acquired by the user, a document created by the user, or the like. In this case, the control unit 21 of the document search apparatus 1 extracts feature words based on the accepted document content itself (step S104) and generates a document vector (step S105).

  Further, the document to be selected from the user in step S101 of the document recommendation process is determined based on the user history information (for example, the document purchase history) stored in advance in the storage unit 22 of the document search device 1 or the like. May be. In that case, the control unit 21 of the document search apparatus 1 extracts feature words from one or more search target documents determined based on the history information (step S104), and generates a document vector (step S105). ).

The document search apparatus 1 can be used alone without being connected to a network. FIG. 13 is a flowchart showing an example of document recommendation processing of the document search apparatus 1a not connected to the network.
The control unit 21 of the document search apparatus 1a receives an input of a search target document from the user (step S601).
Steps S602 to S607 are the same as steps S104 to S109 of the document recommendation process of the document search apparatus 1 shown in FIG.
Then, the control unit 21 of the document search apparatus 1 displays the extracted document with a high unexpectedness score on the document recommendation screen (step S608).

  The preferred embodiments of the document search system 100 and the like according to the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to such examples. It will be apparent to those skilled in the art that various changes or modifications can be conceived within the scope of the technical idea disclosed in the present application, and these naturally belong to the technical scope of the present invention. Understood

DESCRIPTION OF SYMBOLS 1 ......... Document search apparatus 2 (2a, 2b) ......... User terminal 3 ......... Network 5 ......... Category information DB
6 ... Document information DB
21 ......... Control unit 22 ......... Storage unit 23 ......... Media input / output unit 24 ......... Communication control unit 25 ......... Input unit 26 ......... Display unit 27 ......... Peripheral device I / F unit 50 ... …… Category information 56 ……… Category similarity information 57 ……… Document similarity information 60 ……… Document information 100 ……… Document search system

Claims (7)

A document search device that searches for a recommended document that is related to a search target document to be searched,
First storage means for managing and storing feature words for each document of the recommended document;
Second storage means for managing and storing feature words for each category of recommended documents;
Referring to the first storage means, a document similarity calculating means for calculating a document similarity which is a similarity in document units between the target document and the recommended document;
Referring to the second storage means, a classification similarity calculating means for calculating a classification similarity which is a similarity in a classification unit between the target document and the recommended document;
Search means for taking the difference between the document similarity and the classification similarity and extracting the recommended document with reference to the calculated difference value ;
A document search apparatus comprising: Document retrieval system according to claim 1, wherein the extracting those exceeding a threshold which the difference value is set in advance. The document search apparatus according to claim 1 , wherein the search unit obtains a difference by multiplying the classification similarity by a predetermined coefficient. The said search means searches the said recommended document using the said document similarity and / or the said classification similarity which satisfy | fill the preset threshold value. The one of Claim 1 characterized by the above-mentioned. Document retrieval device. A computer-based document search method for searching for a recommended document that is related to a search target document to be searched,
A control unit of the computer,
Document similarity calculation for calculating a document similarity that is a similarity in document units between the target document and the recommended document with reference to the first storage unit that manages and stores the feature word for each document of the recommended document Steps,
Classification similarity calculation for calculating a classification similarity that is a similarity in a classification unit between the target document and the recommended document with reference to a second storage unit that manages and stores the feature word for each classification of the recommended document Steps,
A step of taking a difference between the document similarity and the classification similarity and extracting the recommended document with reference to the calculated difference value ;
A document retrieval method comprising: A program for causing a computer to function as a document search device that searches for a recommended document that is related to a search target document to be searched,
The computer,
First storage means for managing and storing feature words for each document of the recommended document;
Second storage means for managing and storing feature words for each category of recommended documents;
Referring to the first storage means, a document similarity calculating means for calculating a document similarity which is a similarity in document units between the target document and the recommended document;
A classification similarity calculation unit that calculates a classification similarity that is a similarity in a classification unit between the target document and the recommended document with reference to the second storage unit;
Search means for taking the difference between the document similarity and the classification similarity and extracting the recommended document with reference to the calculated difference value ;
Program to function as. A document search system in which a user terminal and a document search device that searches for a recommended document that is related to a search target document to be searched are connected via a network,
The user terminal is
Input receiving means for receiving input of the search target document;
Transmitting means for transmitting the search target document to the document search device;
Display means for receiving and displaying search results from the document search device;
With
The document search device includes:
First storage means for managing and storing feature words for each document of the recommended document;
Second storage means for managing and storing feature words for each category of recommended documents;
Receiving means for receiving the search target document from the user terminal;
Referring to the first storage means, a document similarity calculating means for calculating a document similarity which is a similarity in document units between the target document and the recommended document;
Referring to the second storage means, a classification similarity calculating means for calculating a classification similarity which is a similarity in a classification unit between the target document and the recommended document;
Search means for taking the difference between the document similarity and the classification similarity and extracting the recommended document with reference to the calculated difference value ;
Transmitting means for transmitting the extracted recommended document to the user terminal;
A document retrieval system comprising:

Images