JP2015179385A - Material search device, material search system, material search method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a material search device and the like which search for books/materials that match a user's interest on the basis of a document image with comments.SOLUTION: A material search device 1 stores a feature word database 15 related to appearance frequency of a feature word included in a material to be searched in advance. A document input part 11 takes in an input document 33 image including comments the user brings in. A character recognition/comment extraction part 12 converts printed characters of the input document 33 and a handwritten character 39 into text data and extracts the kind and position of a comment mark 37. A feature word extraction part 13 extracts a feature word included in the text data. A feature word weighting part 14 applies weight, counts the appearance frequency of a feature word in accordance with the kind of comment mark 37, and creates feature data 47. A relevance computation part 16 computes relevance between a search index 31 and the feature data 47, whereas a search result display part 17 presents to a user a highly relevant material to search.

Description

  The present invention relates to a technology for searching for materials, and more particularly, to a material search technology for searching for books and materials having a high degree of association with materials presented by users.

  Conventionally, in order for a user to obtain knowledge, a keyword is input and a related book is searched, or a keyword including the keyword is searched by selecting a keyword in the displayed text. There is.

  For example, Patent Document 1 proposes a system that efficiently presents books including a search keyword input by a user from a terminal and presents them in order of importance.

JP 2013-206388 A

  However, Patent Document 1 has a problem that the user needs to input a keyword and is complicated for the user. In addition, depending on the combination of the input keywords, there is a problem that the number of books cannot be narrowed down or an appropriate book is not presented.

  The present invention has been made in view of the above-described problems, and the object of the present invention is to provide a book / material suitable for the interest / interest of the user related to the contents of the written document brought by the user. An object of the present invention is to provide a material retrieval apparatus that can be presented.

  In order to achieve the above-described object, the first invention extracts a text data by performing a character recognition process on a document image including writing in a material retrieval apparatus that retrieves a material based on the degree of association with feature word data. A text extraction means; a write extraction means for extracting the type and position of the writing; a storage means for storing a search index including a first feature word of the search target material and its importance; a second from the text data; A feature word extracting means for extracting a feature word of the text, and a feature data creating means for calculating the importance of the second feature word using the type and position of the writing and creating the feature word data of the text data And a degree-of-association calculating means for calculating the degree of association between the search index and the feature word data.

  According to the first invention, for a search target material composed of books and materials held by a library or the like, a storage index is stored in advance by the storage means including the importance of the first feature word included in each search target material. In addition, it is possible to obtain the degree of relevance with the second feature word included in the document including the user's writing that has been read, and to present the search target material having a high degree of relevance.

At that time, the writing extraction means extracts the writing position and type written in the document, and the feature data creation means weights the importance of the second feature word at the writing position according to the writing type. By doing so, it becomes possible to present the search target material that matches the user's interest.
The writing type is, for example, underline, marker, box, cross mark, handwritten character, etc. By providing a plurality of writing types, the user can easily express the interest / interest. .

The writing extraction unit performs character recognition processing and adds a recognition result to the text data.
As a result, a user's handwritten memo can be added to the text data, and the second feature word of the memo portion can be used for the search.

It is desirable that the feature data creation means changes the importance of the corresponding second feature word in accordance with the type of writing.
This makes it possible to create feature word data according to the user's interest and interest by increasing or decreasing the importance depending on the type of writing, and present more accurate data.

Further, it is desirable that the feature data creation unit deletes the second feature word corresponding to the type of writing.
Thereby, it becomes possible to remove from the second feature word depending on the type of writing, and it is possible to present materials that more closely match the user's interest.

It is desirable to further comprise index creation means for creating a search index for the search target material.
This makes it possible to update the search index for new search target materials.

It is desirable to further comprise image reading means for reading the document image.
By loading a document that the user has brought with a scanner, for example, it becomes possible to present materials suitable for the content of the document without the user inputting a search keyword, etc., thereby reducing the burden on the user. It becomes possible.
In addition, for example, a document image taken by a user using a camera function of a mobile terminal or the like is loaded into a document retrieval device by an image reading unit via a network such as the Internet, so that a document suitable for the content of the document can be obtained. It becomes possible to present.

As described above, according to the first aspect of the present invention, a user can search for a material suitable for the contents of the document only by taking the user's document as an image into the material searching device without inputting a search keyword into the device. And the burden on the user can be reduced.
Further, by changing the importance of the feature word of the document in accordance with the user's writing, it becomes possible to search for a material that matches the user's interest / interest.

  According to a second aspect of the present invention, there is provided a document retrieval system that retrieves a document based on a degree of association with feature word data, an image reading device that reads a document image including writing, and transmits the read image; A text extracting means for performing processing to extract text data; a writing extracting means for extracting the type and position of the writing; an extracting device for transmitting the extracted data; and a first characteristic word of the search target material; Storage means for storing a search index including the degree of importance, feature word extraction means for extracting a second feature word from the text data, and the type and position of the writing, the second feature word Feature data creating means for calculating importance and creating the feature word data of the text data, the search index, and the feature word data , A relevance calculation means for calculating a degree of association with a document retrieval system characterized by comprising a server, a with a.

According to the second invention, it is possible to present a material suitable for the contents of the document only by allowing the user to read the document image of the user without inputting the search keyword into the system. The burden can be reduced.
Further, by changing the importance of the feature word of the document in accordance with the user's writing, it becomes possible to search for a material that matches the user's interest / interest.

  A third invention is a material retrieval method performed by a material retrieval device that retrieves materials based on the degree of association with feature word data, and performs text recognition processing on a document image including writing to extract text data A step of extracting, a writing extraction step of extracting the type and position of the writing, a storing step of storing a first feature word of the search target material and a search index including its importance, and a second feature from the text data A feature word extracting step for extracting a word; a feature data creating step for calculating the importance of the second feature word using the type and position of the writing; and creating the feature word data of the text data; And a relevance level calculating step for calculating a relevance level between the search index and the feature word data.

  A fourth invention is a program for causing a computer to function as a material retrieval device that retrieves material based on the degree of association with feature word data, wherein the computer performs character recognition processing on a document image including writing. Text extracting means for extracting the applied text data, write extracting means for extracting the type and position of the writing, storage means for storing a first feature word of the search target material and a search index including its importance, Importance of the second feature word is calculated using feature word extraction means for extracting a second feature word from the text data, and the type and position of the writing, and the feature word data of the text data is The feature data creating means to be created and the relevance calculation means for calculating the relevance between the search index and the feature word data Which is the program.

  By installing the program according to the fourth invention in a general-purpose computer, the material search apparatus according to the first invention can be obtained and the material search method according to the third invention can be executed.

  With the material retrieval apparatus of the present invention, it becomes possible to present books and materials related to the contents of written documents brought by users.

1 is a block diagram showing a hardware configuration of a material retrieval apparatus 1 according to the present embodiment. The block diagram which shows the function structure of the material search device 1 which concerns on this embodiment. The figure which shows the structural example of the feature word database 15 of search object material The figure which shows the example of the input document 33 The flowchart which shows the flow of a process of the material search device 1 which concerns on this embodiment. The figure which shows the example of the write mark data 41 The figure which shows the example of the feature word extracted from the input document 33 The figure which shows the example of the weighting magnification 45 of a writing mark The figure which shows the example of the characteristic data of the input document 33 Diagram explaining the degree of association between feature word data and search index of search target material The figure which shows the example of the search result output screen 55 The figure which shows the system configuration example of the material search system 10

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram illustrating a hardware configuration example of the material retrieval apparatus 1. As shown in FIG. 1, the material retrieval apparatus 1 can be configured by a computer system. The configuration in FIG. 1 is merely an example, and various configurations can be adopted depending on the application and purpose.

  In the material retrieval apparatus 1, for example, the control unit 21, the storage unit 22, the media input / output unit 23, the communication control unit 24, the input unit 25, the display unit 26, the peripheral device I / F (interface) unit 27, etc. Connected and configured.

  The control unit 21 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

The CPU calls and executes a program stored in the ROM, the storage unit 22 and the like in a work memory area on the RAM, drives and controls each device connected via the bus 28, and realizes processing performed by the computer.
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 program, data, and the like loaded from the storage unit 22, ROM, storage medium, and the like, and includes a work area used by the control unit 21 to perform various processes.

  The storage unit 22 stores a program executed by the control unit 21, data necessary for program execution, an OS (Operating System), and the like. The storage unit 22 stores a feature word database 15 and a dictionary database 18 of a search target material, which will be described later, and a program and data of the material search device 1 of the present embodiment.

The media input / output unit 23 (drive device) inputs / outputs data, for example, media such as a CD drive (-ROM, -R, -RW, etc.), DVD drive (-ROM, -R, -RW, etc.) Has input / output devices.
The communication control unit 24 includes a communication control device, a communication port, and the like, and performs communication control with other devices via a network. The network may be wired or wireless.

The input unit 25 inputs data and includes, for example, a pointing device such as a keyboard, a mouse, and a touch panel, and an input device such as a numeric keypad.
The display unit 26 is a display device such as a CRT monitor or a liquid crystal panel, and the search result is displayed on the display unit 26 in the material search device 1 of the present embodiment.

The peripheral device I / F (interface) unit 27 is a port for connecting peripheral devices, and is configured by USB, IEEE1394, RS-232C, or the like, and the connection form may be wired or wireless.
For example, a scanner is connected via the peripheral device I / F unit 27, and image input data of a document brought by the user can be captured.
The bus 28 is a path that mediates transmission / reception of control signals, data signals, and the like between the devices.

  In addition, the document search apparatus 1 may include a camera (not shown) for inputting image data, and a scanner (not shown) may be connected to the peripheral device I / F (interface) unit 27.

FIG. 2 is a block diagram illustrating a functional configuration example of the material search apparatus 1 according to the embodiment of the present invention.
The document search device 1 includes a document input unit 11, a character recognition / write extraction unit 12, a feature word extraction unit 13, a feature word weighting unit 14, a feature word database 15 of a search target material, a relevance calculation unit 16, and a search result display unit. 17 and a dictionary database 18 and the like.

The feature word database 15 of the search target material is a database that stores, for example, a search index for searching for books and materials held in libraries such as universities by the material search device 1 of the present embodiment.
As will be described in detail later, the search index is data relating to the importance of words (characteristic words) that are characteristic of each book and document. The dictionary database 18 is used in advance from bibliographic data and full text of each book and document. Create by extracting feature words and determining their importance.

  The dictionary database 18 is, for example, a database that stores headings recorded in several dictionaries, and is used for extracting feature words. The part of speech of the headline stored in the dictionary database 18 may be only a noun, but other parts of speech (verbs, adjectives, etc.) may be stored and used.

The document input unit 11 can be configured by, for example, a scanner or a camera.
The document input unit 11 captures a document brought in by the user as image data.

Documents are, for example, university syllabuses and resumes, related materials, lecture notes, one page of books, newspapers and magazine articles, etc., which are printed matter of printed characters, but handwritten by users There may be.
The writing is, for example, marking with an underline or a marker, a box, a handwritten character, an X mark for excluding unnecessary portions, and the like.

The character recognition / write extraction unit 12 executes character recognition processing on the image data of the document captured by the document search unit 1 by the document input unit 11 and converts it into text data. Extract type and position.
When the handwritten character is written, a character recognition process is executed for the handwritten character, and the handwritten character is converted into text data.

The feature word extraction unit 13 extracts feature words from the text data converted by the character recognition / writing extraction unit 12 with reference to the dictionary database 18.
The feature words are, for example, noun words and unknown words that are not in the dictionary database 18.

  The feature word extraction unit 13 first performs morphological analysis on the text data, searches the dictionary database 18 for the noun part of the text data, and matches words (for example, “history”, “female”, “feminism”, “Japan” Etc.) as feature words. In addition, when it is determined as a noun in the morphological analysis and there is no matching word in the dictionary database 18 (for example, “Abenomics” or the like), it is added to the feature word as an unknown word.

Next, the feature word weighting unit 14 obtains the importance for the feature word extracted by the feature word extraction unit 13 and weights the importance according to the content of the writing by the user.
As the importance, for example, various importance calculation methods such as the appearance frequency of feature words and TF / IDF (Term Frequency / Inverse Document Frequency) values can be used.

When the appearance frequency is used for importance, for example, in the case of underlining, marking, enclosing, or writing of handwritten characters, it is determined that the user is writing as an important part or an interesting / interested part, and that part For example, the appearance frequency is counted by multiplying the weight of a positive coefficient, for example, twice.
On the other hand, the feature words included in the portion where the x mark is written in the document may not be counted.

  By the weighting of the importance of the feature word for the written part by the feature word weighting unit 14, it is possible to search for an appropriate material that matches the user's interest.

  The degree-of-relevance calculation unit 16 stores the feature data (feature words and their importance data) about the user's document obtained by the feature word weighting unit 14 and the feature word database 15 of the search target material. Calculate the relevance of the search target material.

  The search result display unit 17 displays the name of the search target material having a high relevance level based on the relevance level calculated by the relevance level calculation unit 16.

Next, the feature word database 15 of the search target material will be described with reference to FIG.
FIG. 3 is a diagram illustrating a configuration example of the feature word database 15 of the search target material.

The feature word database 15 of the search target material includes a search index 31 for each book or material to be searched. P search indexes 31-1 to 31 -P for the P search target materials are stored in the feature word database 15.
The search index 31 is created / added as the number of collections / materials such as libraries increases, and the feature word database 15 is updated.

The search index 31 includes, for example, a search target material ID, a feature word heading, and its importance.
As shown in FIG. 3, for example, the search index 31-1 for the search target material ID “1” includes feature words such as “female”, “profession”, “culture”, “Japan”, and the like. Consists of importance.

  The importance of feature words in the search index 31 is based on the appearance frequency of feature words included in the bibliographic data of the material to be searched and the full text of the material, but is weighted by, for example, the TF / IDF method. It is desirable to be.

  The TF / IDF method is a well-known technique and will not be described in detail. However, the TF / IDF method is a value (TF) related to the ratio of the appearance frequency of a specific word to the appearance frequency of all words included in the specific document. And the appearance frequency based on the value (IDF) relating to the ratio of the number of documents including the specific word in the total number of documents. According to the TF / IDF method, for example, the appearance frequency of words that frequently appear in any document such as “this” and “that” is suppressed, and the appearance frequency of words that frequently appear only in a specific document increases. .

FIG. 4 is a diagram illustrating an example of the input document 33 that the user inputs to the material search apparatus 1.
In the input document 33, an article by the print character 35 is printed, and a writing mark 37 and a handwritten character 39 written by the user by hand are drawn.

In the input document example 33 of FIG. 4, articles A to D are printed, and the user has written writing marks 37 a to 37 d and handwritten characters 39.
The writing mark 37 is, for example, a marker 37a (writing mark a), an underline 37b (writing mark b), an enclosure 37c (writing mark c), an X mark 37d (writing mark d), or the like.

The marking 37a, the underline 37b, and the enclosure 37c are written in a portion that the user considers important, and weighting may be performed so that the importance of the feature word included in the portion is increased.
Further, the x mark 37d is written in a portion that the user thinks is unnecessary, and the feature words included in the portion may be excluded.

  The handwritten character 39 is considered to be a sentence or sentence that the user considers important, and the feature words included therein are important and weighted so as to increase the importance.

Next, the flow of processing of the material retrieval apparatus 1 according to this embodiment will be described.
FIG. 5 is a flowchart showing a processing flow of the material search apparatus 1.

First, the control unit 21 of the material retrieval apparatus 1 captures an image of the input document 33 (step 101).
For example, the image of the input document 33 is read by a scanner connected to the peripheral device I / F unit 27 and stored in the storage unit 22.

  The method of capturing the input document 33 is not limited to the scanner. For example, the user can photograph the input document 33 with a camera such as a portable terminal and send the image to the document retrieval apparatus 1 via the network. 24 may be received and stored in the storage unit 22.

Next, the control unit 21 performs character recognition processing on the captured image data, and converts the print characters 35 and handwritten characters 39 of the input document 33 into text data (step 102).
For the character recognition process, a known technique such as OCR (Optical Character Recognition) may be used.
The text data extracted by the character recognition process for handwritten characters may be attached with a flag indicating that it is a handwritten character.

Next, the control unit 21 extracts the writing mark 37 written by the user, and obtains the type and position (step 103).
Extraction from the image of the input document 33 is performed assuming the type of writing (marking 37a, underline 37b, box 37c, x mark 37d).

For example, in the case of the marking 37a, a substantially rectangular shape overlapping the text portion is extracted.
In the case of the underline 37b, a substantially straight line shape that does not overlap the text is extracted.
In the case of the enclosure 37c, the shape of a closed curve other than characters is extracted.
Further, in the case of the x mark 37d, the shape of two diagonally intersecting straight lines is extracted.

The type and position data of the write mark 37 extracted in step 103 is stored in the storage unit 22 as write mark data 41.
FIG. 6 is a diagram illustrating a configuration example of the write mark data 41.
The write mark data 41 includes an input document ID that is an identification number of the input document 33, a mark No that indicates the identification number of writing included in the input document 33, a write mark ID that indicates the type of the write mark 37, and the write It consists of mark position data.

The position data is, for example, two-dimensional coordinates with the upper left corner of the input document 33 as the origin.
In the case of the marking 37a, the coordinates of the substantially rectangular diagonal vertices, in the case of the underline 37b, the coordinates of both ends of the straight line, and in the case of the enclosure 37c, the (closed x curve, minimum x coordinate) and (maximum x coordinate, In the case of (maximum y coordinate) and x mark 37d, two lines (minimum x coordinate, minimum y coordinate) and (maximum x coordinate, maximum y coordinate) can be used as position data.

Next, the control unit 21 performs morphological analysis on the text data (step 104).
That is, the text data is divided into meaningful words and the part of speech is identified using the dictionary database 18.

Next, the control unit 21 extracts nouns in words subjected to morphological analysis and unknown words that cannot be searched in the dictionary database 18 as feature words (step 105).
FIG. 7 is a diagram illustrating an example of the feature word 43 extracted from the text data of the input document 33.
As shown in FIG. 7, feature words included in the text data are extracted.
In addition, a flag indicating that is added to the feature word extracted from the text data of the handwritten character 39 portion.

Next, the control unit 21 counts the weighting importance of each feature word and creates feature data of the input document 33 (step 106).
The weighting importance is obtained by counting the appearance frequency of feature words extracted from the text data based on the writing mark data 41 shown in FIG. 6 and the weighting magnification 45 shown in FIG.

FIG. 8 is a diagram illustrating an example of the weighting magnification 45.
For example, when the writing mark 37 is the marking 37a and the underline 37b, the appearance frequency of the feature word included in the position of the marking 37a and the underline 37b is 2.0 times, and when the writing mark 37 is the surrounding 37c, it is included in the surrounding position. The frequency of occurrence of feature words is counted as 1.7 times.
When the writing mark 37 is an x mark 37d, the weighting magnification of the feature word included in the range of the position of the x mark is set to 0 so as not to count.
Furthermore, in the case of the handwritten character 39, for example, the appearance frequency of the feature word flagged as the handwritten character is weighted by 2.5 times and counted.

FIG. 9 is a diagram illustrating an example of the feature word data 47 created in step 106. The case where the appearance frequency is used as the importance is shown.
The feature data 47 of the input document 33 includes feature words and weighted appearance frequency (weighted importance).
By assigning weights, the value of the weighted appearance frequency becomes an increased or decreased numerical value, unlike the appearance frequency of the feature word actually included in the input document 33 (the numerical value in parentheses).

  As described above, the feature data 47 in the material retrieval apparatus 1 according to the present embodiment is weighted according to the user's writing marks 37 and handwritten characters 39, and the feature words that the user considers interesting / interesting and important are displayed. It becomes data that appears more accurately, and more accurate material search becomes possible.

Next, the control unit 21 calculates the degree of association between the feature data 47 of the input document 33 and the search index 31 of each search target material in the feature word database 15 (step 107).
For the calculation of the degree of association, for example, a cosine similarity that is a known technique may be used.

FIG. 10 is a diagram illustrating the degree of association between the feature word data 47 and the search index 31 of the search target material.
The cosine similarity is an angle θ formed by the vector 51 of the search index 31 and the vector 53 of the feature data 47. The smaller the angle θ, the higher the similarity, that is, the higher the degree of association between the two vectors.

In FIG. 10, in order to simplify the explanation, a three-dimensional vector for three types of feature words is described as an example, but the elements of the vectors 51 and 53 are the search target material and the input document 33. Is the importance (weighted) of a plurality of feature words included in.
The magnitude of the angle θ indicating the relevance is obtained by calculating the inner product of the vector 51 of the search index 31 and the vector 53 of the feature data 47 of the input document 33.

  Next, the control unit 21 compares the relevance values calculated in step 107, searches for material names based on the identification numbers of the search target materials having high relevance levels, and displays them on the display unit 26 (steps). 108).

FIG. 11 is a diagram illustrating an example of the search result output screen 55.
On the display unit 26 of the material retrieval apparatus 1, documents, books, names of materials, etc. that are highly relevant to the input document 33 presented by the user are displayed.

  As described above, the material retrieval apparatus 1 according to the present embodiment reads the document 33 brought by the user with the scanner, thereby obtaining the retrieval target material having a high degree of association with the document with the weighted appearance frequency of the feature word. Can be retrieved and displayed as a scale.

  In addition, the material retrieval apparatus 1 according to the present embodiment assigns a weight according to the type of the writing mark 37 to the appearance frequency of the feature word of the portion by the user writing the writing mark 37 or the handwritten character 39 in the document. This makes it possible to search for a search target material that more matches the user's interests / interests.

  In the above description, the material retrieval apparatus 1 according to the present embodiment has been described as a single apparatus. However, the material retrieval apparatus 1 includes an image reading apparatus 110 such as a scanner, an extraction apparatus 120 such as an OCR, and a server 130 that performs search processing. The document search system 10 may be configured.

FIG. 12 is a diagram illustrating a system configuration example of the material search system 10.
As shown in FIG. 12, the material search system 10 has a configuration in which an image reading device 110, an extraction device 120, and a server 130 are communicably connected via a network 140, for example.

The image reading device 110 can be constituted by a scanner, for example, and reads the document 33 including writings brought by the user.
The read document image data is sent to the extraction device 120 via the network 140.

The extraction device 120 can be configured by an OCR device, for example.
The extracting device 120 receives the document image data, performs recognition processing of the print character 35 and recognition processing of the handwritten character 39, creates text data, extracts the writing mark 37 written in the document 33, and extracts the writing mark Data 41 is created.
The created text data and write mark data 41 are sent to the server 3 via the network 140.

The server 3 includes a feature word database 15 and a dictionary database 18 created from search target materials.
The server 3 can be composed of a general-purpose computer or the like, and executes the processing of steps 104 to 108 in the flowchart of FIG.

  That is, the server 3 performs a process of extracting feature words from the text data received from the extracting device 120 using the dictionary database 18 (steps 104 and 105), and based on the writing mark data 41 received from the extracting device 120. Then, the weighted importance of the extracted feature word is calculated to create feature data (step 106), the degree of association between each search index in the feature word database 15 and the feature data is calculated (step 107), A high search target material is presented to the user (step 108).

  In the above description, the user sends an image of the document 33 from a portable terminal, personal computer, or the like to the material retrieval system 10 via a network such as the Internet, and the sent document image is processed by the extraction device 120 and the server 130. The search result may be sent to a portable terminal or a personal computer via a network such as the Internet, and the search result may be displayed on the display unit.

  In the above description, nouns and unknown words are used as feature words. However, other parts of speech may be used.

Further, the appearance frequency of the feature word in the feature word database 15 of the search target material is preferably weighted by the TF / IDF method, but it indicates whether or not the feature word has appeared without using the appearance frequency of the feature word 2 It may be expressed by a value vector.
Moreover, you may make it use the vector which makes the element co-occurrence frequency (word N-gram) of the feature word which adjoins instead of the appearance frequency of a feature word.

  As mentioned above, although embodiment of this invention was described referring an accompanying drawing, the technical scope of this invention is not influenced by embodiment mentioned above. It is obvious for those skilled in the art that various modifications or modifications can be conceived within the scope of the technical idea described in the claims. It is understood that it belongs.

DESCRIPTION OF SYMBOLS 1 ......... Material search device 10 ......... Material search system 11 ......... Document input part 12 ......... Character recognition and writing extraction part 13 ......... Feature word extraction part 14 ......... Feature word weighting part 15 ... ... Characteristic database of search target material 16 ......... Relevance calculation unit 17 ......... Search result display unit 18 ......... Dictionary database 31 ......... Search index 33 ......... Input document 35 ......... Print characters 37 ......... Writing mark 39 ......... Handwritten character 41 ......... Writing mark data 45 ......... Weighted magnification 47 ......... Characteristic data of input document 33

Claims (9)

In a material retrieval device that retrieves materials based on the degree of association with feature word data,
Text extraction means for extracting text data by performing character recognition processing on a document image including writing;
Write extraction means for extracting the type and position of the write;
Storage means for storing a search index including a first feature word of the search target material and its importance;
Feature word extraction means for extracting a second feature word from the text data;
Calculating the importance of the second feature word using the type and position of writing, and feature data creating means for creating the feature word data of the text data;
Relevance calculation means for calculating relevance between the search index and the feature word data;
A material retrieval apparatus comprising:   The material retrieval apparatus according to claim 1, wherein the writing extraction unit performs a character recognition process and adds a recognition result to the text data.   The material search apparatus according to claim 1, wherein the feature data creation unit changes the importance of the second feature word in accordance with the type of writing.   4. The material retrieval apparatus according to claim 1, wherein the feature data creating unit deletes the second feature word corresponding to the type of writing.   5. The material retrieval apparatus according to claim 1, further comprising index creation means for creating a search index for the retrieval target material.   6. The material retrieval apparatus according to claim 1, further comprising image reading means for reading the document image. In a material retrieval system that retrieves materials based on the degree of association with feature word data,
An image reading device for reading a document image including writing and transmitting the read image;
A text extraction means for performing character recognition processing on the document image and extracting text data; and a writing extraction means for extracting the type and position of the writing; and an extraction device for transmitting the extracted data;
A storage means for storing a first feature word of the search target material and a search index including its importance; a feature word extraction means for extracting a second feature word from the text data; and the type and position of the writing Calculating the degree of importance of the second feature word by using the feature data creating means for creating the feature word data of the text data, and the relationship for calculating the degree of association between the search index and the feature word data A server comprising: a degree calculation means;
A material retrieval system comprising: A material retrieval method that is performed by a material retrieval device that retrieves materials based on the degree of association with feature word data,
A text extraction step of extracting text data by performing character recognition processing on a document image including writing;
A write extraction step for extracting the type and position of the write;
A storage step of storing a search index including a first feature word of the search target material and its importance;
A feature word extraction step of extracting a second feature word from the text data;
Calculating a degree of importance of the second feature word using the type and position of writing, and a feature data creating step of creating the feature word data of the text data;
A relevance calculating step for calculating a relevance between the search index and the feature word data;
A material retrieval method characterized by including: A program for causing a computer to function as a material retrieval device that retrieves material based on the degree of association with feature word data,
The computer,
Text extraction means for extracting text data by performing character recognition processing on a document image including writing;
Write extraction means for extracting the type and position of the write;
Storage means for storing a search index including a first feature word of the search target material and its importance;
Feature word extraction means for extracting a second feature word from the text data;
Calculating the importance of the second feature word using the type and position of writing, and feature data creating means for creating the feature word data of the text data;
Relevance calculation means for calculating relevance between the search index and the feature word data;
Program to function as.

Images