JP2005227975A - Document search device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To focus on a role of a unique expression in a text for performing search with less noise. <P>SOLUTION: A unique expression extraction part 12 extracts the unique expression in the text and stores a document ID, a unique expression (appearance form), a unique expression (standard form), a category, an offset, length, and a role in a unique expression storage part 13. A search condition input part 14 inputs a search condition designating arbitrary one from a search character string, a category, an instruction expression search flag, a role, and a document ID. Using the search condition in the search condition storage part 15, a search part 16 searches for the unique expression record stored in the unique expression storage part 13 and stores the search result in a search result storage part 17. An output part 18 emphasizes and displays a part of the search result in a document by referring to the search result of the search result storage part 17 and document contents of the document storage part 11 to output the search result part. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a technique for providing a search result with less search noise that matches a searcher's intention.

  Conventional highlighting emphasizes a character string that matches a specified condition, for example, highlights a character string in a document that matches a character string (word) in a search condition ( For example, Google, trademark, ConceptBase Search, trademark). In addition, a screen having a function of displaying a list of compatible locations on a separate screen like a table of contents and jumping from the table of contents screen to the appearance location is also suitable (for example, DocuWorks 6.0, trademark).

By the way, the conventional search is a search / highlight display for the character string on the surface layer, and it has been difficult to narrow down to a search that matches the user's intention. It is desired to perform a search that matches the user's intention in consideration of the metadata of the character string.
(As a problem of the prior art, highlighting for character strings on the surface layer is not highlighting for metadata that does not appear on the surface layer such as the role played by the character strings.
In addition, the instruction expression (this, the company, etc.) cannot be determined to be suitable for the search condition by the simple comparison, and has not been highlighted. )

  Patent documents related to this invention include the following.

  Patent Document 1 discloses storing organization data including information related to an organization name and organization structure, and human data including information related to a name and an affiliated organization, together with the past change history, in the personnel affairs DB. The document data is stored in the document DB together with the attribute information. When document attribute information is input as a search condition for searching stored document data, the input search condition is corrected to a valid search condition based on the change history stored in the personnel database, or limit. Even if you search by past organization name, you can search appropriately. However, this proposal performs a search based on attribute information, and does not extract a specific expression or consider its role.

  In Patent Document 2, when creating a word index from text, the part of speech and attribute information of the word as well as the information of the word headline and the text including the word as well as the part of speech and attribute information are registered. , Search using an attribute (for example, a specific expression) as a search condition is possible, and excessive search is suppressed particularly when an ambiguous word is used as a search key. In addition, based on the part of speech of the word and the added attribute, the word with ambiguousness is expanded and registered in the index, and a word index in which the ambiguous word is correctly expanded is generated. This makes it possible to search for synonymous words even if they are different from the search key, reduce search omissions, and reduce the search load compared to the case where the user himself / herself expands the search key into synonyms and searches. Can do.

  However, this proposal does not focus on the role of specific expressions.

  Patent Document 3 extracts a proper noun from the description of a document to be registered in a database, adds the extracted proper noun as additional information of the document to be registered, and registers it in the database. When an input for searching the database is received, the database is searched by the received input, and the degree of fitness with the input for searching the retrieved document is calculated as a score. A score is added to a document obtained every time additional information included in the document matches, and a proper noun score is calculated. The calculated proper noun score is displayed together with the proper noun.

This proposal also does not focus on the role of specific expressions.
JP-A-10-27180 Japanese Patent Laid-Open No. 11-39347 JP 2001-273328 A

  The present invention has been made in consideration of the above circumstances, and an object thereof is to provide a search technique capable of obtaining a search result that matches a user's intention.

  In a specific configuration example of the present invention, in order to achieve the above-described object, when searching for a specific expression existing in a document, a role (subject, object, attribute) in the sentence of the specific expression is specified. Search for objects that match the proper expression and role, and realize search with less search noise. In addition, it enables detection and display of parts necessary for understanding the text content, and supports reading comprehension.

  The present invention will be further described.

  According to one aspect of the present invention, in order to achieve the above object, in a document search apparatus: a document registration unit for registering a document; a document storage unit for storing a registered document; and a storage in the document storage unit A search means for searching for a document that is stored based on the specific expression and the role of the specific expression in the sentence; and a search result output means for outputting a search result by the search means.

  In this configuration, it is possible to perform a search with less noise by paying attention to the role of the unique expression.

  A proper expression (also called a proper name) is an important expression unit in a sentence, such as a person name, an organization name, a place name, a currency, and a date.

  In this configuration, the search result output means highlights and presents the unique expression searched by the search means. The corresponding specific expression itself may be highlighted, or a sentence unit or paragraph unit including the corresponding specific expression may be highlighted. Instead of displaying the entire document including the corresponding specific expression, it is also possible to display the summary or display the corresponding sentence or paragraph. When displaying the summary, the corresponding sentence, and the corresponding paragraph, the corresponding specific expression may be highlighted.

  In addition, the search means outputs a primary search result based on the specific expression, marks only the search result that matches the role, makes the marked search result the final search result, and outputs the search result. You may make it output by an output means. Specifically, an “ON” mark is given based on the specific expression, and “ON” is reset to “OFF” for those that do not match the role. Of course, it is not limited to this. In this case, when the designation of the role is reset, a mark that does not match the above-described role may be added.

  The search means may perform a search by designating a category of the specific expression. The category is, for example, an organization, a person name, or a place name, but may be further broken down.

  The search means may perform a search by designating a character string of the unique expression.

  Also, the roles in the above sentence are, for example, a role as a subject of action and state, a role as an object of action and state, and other roles.

  The role in the sentence is determined by, for example, particles and phrases corresponding to particles, but is not limited thereto.

  According to another aspect of the present invention, in order to achieve the above-described object, the document search apparatus includes: a document unit for inputting a document; a document input by the document input unit; Search means for searching based on the role in the sentence; and search result output means for outputting the search result by the search means are provided.

  Even in this configuration, it is possible to perform a search with less noise by paying attention to the role of the document of the unique expression.

  The present invention can be realized not only as an apparatus or a system but also as a method. Of course, a part of the invention can be configured as software. Of course, software products used to cause a computer to execute such software are also included in the technical scope of the present invention.

  These and other aspects of the invention are set forth in the appended claims and will be described in detail below with reference to examples.

  According to the present invention, it is possible to perform a search with less noise by paying attention to the role in the sentence of the specific expression.

  Examples of the present invention will be described below.

  FIG. 1 shows an overall embodiment of a document search apparatus to which the present invention is applied. In this figure, the document search apparatus includes a document registration unit 10, a document storage unit 11, a specific expression extraction unit 12, a specific expression storage. A unit 13, a search condition input unit 14, a search condition storage unit 15, a search unit 16, a search result storage unit 17, an output unit 18, and the like. A part of each functional block shown in the figure is actually mounted on the computer 100 as a computer program.

  The document registration unit 10 inputs one or a plurality of documents. The document specified by the user may be input, the document specified by the destination may be received and input as it is, or the document is classified using keywords, word vectors, etc., and a predetermined classification group You may enter the document. The document input by the document registration unit 10 is stored in the document storage unit 11. The stored document is, for example, as shown in FIG. 8, and a document ID is assigned. Note that the company name and the product name included in the sentence of the case are trademarks, and the person name is fictitious rather than displaying a real person. Documents may be managed separately from document attributes and document content.

  The specific expression extraction unit 12 extracts a specific expression (also referred to as a specific name) in the document. The proper expression is an important expression unit in a sentence such as a person name, an organization name, a place name, a currency, and a date. For example, as shown in FIG. 2, the specific expression extraction unit 12 includes a morpheme analysis unit 121, a morpheme analysis dictionary storage unit 122, a rule application unit 123, a rule storage unit 124, and the like.

  The input of the specific expression extraction unit 12 is as shown in FIG. 3, for example, and the morpheme analysis dictionary in the morpheme analysis dictionary storage unit 122 has an entry as shown in FIG. The morpheme analysis unit 121 obtains an analysis result as shown in FIG. 5 from the input example (FIG. 3) using the morpheme analysis dictionary. In FIG. 5, “/” indicates a delimiter between morphemes, and a portion surrounded by “<” and “>” indicates a part of speech. In the figure, the starting position and length are omitted. The morpheme analysis result is input to the rule application unit 123, and a specific expression is extracted with reference to the extraction rule in the rule storage unit 124. FIG. 6 shows an example of the extraction rule. For example, “last name” and “first name” are combined by the rule number 5 to give the category “PERSON”. The extraction result is as shown in FIG. In this example, each unique expression is extracted and given a category such as <ORGANIZATION>, <PERSON>, <CURRENT>, <DATE>, <PLACE>, and the like.

  Further, the specific expression extraction process will be described. FIG. 9 shows the process of the specific expression extraction unit 12. In this figure, individual target documents are sequentially extracted, and the specific expression is extracted from the document contents (shown in FIG. 8) (S10). . The extracted specific expression information is stored in the specific expression storage unit 13 (S11). The specific expression information includes, for example, a document ID, a specific expression (appearance form), a specific expression (standard form), a category, an offset, and a length, but is not limited thereto. The specific expression (standard form) can be determined by searching the standard title table 125 of FIG. If the role table (role / particle (equivalent phrase) correspondence table) 127 shown in FIG. 12 is followed by a role related particle (equivalent phrase) following the extracted specific expression, it corresponds to the role of the specific expression. The role is also registered (S12). When the instruction expression of the instruction expression table 126 shown in FIG. 11 is extracted, it is also stored in the specific expression storage unit 13 (S13). The category is PRONOUN. A specific expression having a category specified in the instruction expression table is searched from the specific expression storage unit 13, and the record number of the closest specific expression is registered as a reference destination. The above processing is executed for each document (S14). The standard notation table 125 (FIG. 10), the instruction expression table 126 (FIG. 11), and the role table 127 (FIG. 12) are stored in the storage unit 19 of the computer 100.

  The extracted specific expressions (including the instruction expression) are stored in the specific expression storage unit 13 as shown in FIG. In this example, a record number, a document ID to which it belongs, a specific expression (appearance form), a specific expression (standard form), a category, an offset, a length, a role, a reference destination, and the like are given to the extracted specific expression. .

  Returning to FIG. The search condition input unit 14 is input by a user using, for example, a predetermined graphical user interface (GUI). A search condition fixed in advance may be designated by an administrator or the like. In the following example, “search character string”, “category”, “instruction expression search flag” (specifying whether to search using the instruction expression), “role”, “document ID” are input from the search condition input unit 14. A search condition designating any one of them is input (FIGS. 16 to 22). The “document ID” may be a document name or a search using other document attributes. The search condition is stored in the search condition storage unit 15. The search unit 16 searches for the record of the specific expression stored in the specific expression storage unit 13 using the search condition of the search condition storage unit 15, and stores the search result in the search result storage unit 17.

  The output unit 18 refers to the search result of the search result storage unit 17 and the document content of the document storage unit 11 and outputs the document with the search result portion (the searched unique expression portion) highlighted (FIG. 16). To FIG. 22). Of course, it is also possible to automatically generate a summary instead of the full text of the corresponding document, and to output the highlighted unique expression portion.

  Furthermore, the process of the search part 16 is demonstrated with reference to FIG. 14, the search condition input unit 14 registers the search condition set by the user through the GUI, for example, in the search condition storage unit 15 (S20). The search unit 16 searches for a document or an appearance location in the document that matches the search character string, category, and document ID in the search condition storage unit 15, and displays the result in the search result storage unit 17 as its record number. The flag (ON during registration) is stored as shown in FIG. 15 (S21). If no document ID is specified in the search condition of the search condition storage unit 15, the search is executed for all the records in the specific expression storage unit 13. If the document ID is designated, a search is executed for the record group of the document ID. Also, if the instruction expression search flag is ON in the search condition of the search condition storage unit 15, the record of the instruction expression having the record number of the searched unique expression, the category of PRONOUN, and the same document ID is also searched. The result is registered in the search result storage unit 17. If a role is specified in the search condition of the search condition storage unit 15, the display flag of the record that is not the specified role is further turned OFF in the searched unique expression record (display as described above). The flag is registered as ON). It is also possible to immediately compare a search result when the role is specified by switching ON / OFF of the display flag of the search result record in the search result storage unit 17 and a search result when the role is not specified.

  Further, the search process will be described using a search example.

[Search by category]
FIG. 16A uses a specific expression of a specific category (in this example, “ORGANIZATION”) as a search condition. The search result is as shown in FIG. 16B, and since the role is not specified, the display flag remains “ON” at the beginning, record number 1 (document ID = 1, see FIG. 13), record Number 6 (document ID = 2), record number 7 (document ID = 2), record number 8 (document ID = 3), record number 9 (document ID = 3), record number 10 (document ID = 3), record The number 13 (document ID = 3) is retrieved and displayed. As shown in FIG. 16C, the document is displayed with the specific expression of the category of the organization name highlighted. In the figure, the emphasized portion is underlined, but various changes such as changing the color or attaching a blink attribute can be employed. You may attach marks, such as an arrow.

[Search by category and role]
The example of FIG. 17 includes a role in a search condition in addition to a specific category. In this example, the category is “ORGANIZATION” (FIG. 17A), and the role is “subject”. As in the case of FIG. 16, record numbers 1, 6, 7, 8, 9, 10, and 13 are included in the search results. However, since the role is designated as “subject”, the display flag is reset to “OFF” for the record numbers 7, 10 and 13 of the other roles. Therefore, in the output example of FIG. 17C, the highlighting of a part of the unique expressions is reset as compared with FIG.

[Search by specifying character string and role]
In the example of FIG. 18, the search condition includes the character string (keyword) “Fuji Xerox” (trademark), the role “subject”, and the document ID = 2 (FIG. 18A). The search result is as shown in FIG. 18B, and the output result is as shown in FIG.

[Search by specifying all categories]
In the example of FIG. 19, the search conditions (see FIG. 19A) are all categories. In this case, as shown in FIG. 19B, all the unique expressions are highlighted.

[Search by specifying a character string]
In the example of FIG. 20, the search condition is only the designation of a character string (see FIG. 20A). In this case, as shown in FIG. 20B, the character string (in this example, “Fuji Xerox Co., Ltd.” or “Fuji Xerox”. Appearance form. Trademark) is highlighted.

[Search by category and document ID]
In the example of FIG. 21, a category and a document ID are designated as search conditions (see FIG. 21A). In this case, as shown in FIG. 21B, the designated document is displayed with the specific expression of the designated category highlighted.

[Search by specifying character string and document ID]
In the example of FIG. 22, a character string and a document ID are specified as a search condition (see FIG. 22A). In this case, as shown in FIG. 22B, the designated document is displayed with the designated character string (appearance form) highlighted.

  Note that when the instruction expression search is designated and the instruction expression search flag is turned ON, the instruction expression portion is also highlighted.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the invention. For example, in the above example, the document search device is mounted on the computer 100, but the document search device may be mounted in the form of a client server as shown in FIGS. In the example of FIG. 23, the search condition input unit 14 and the output unit 18 are arranged in the client computer 200, and the rest are arranged in the server computer 300. In the example of FIG. 24, the document registration unit 10 is also arranged in the client computer 200. Also, as shown in FIG. 25, the user may specify and input a document via the document input unit 20 and immediately extract and output the unique name. For example, a program for a document search apparatus may be installed as a plug-in for a predetermined document processing application so as to extract a unique name of a document being processed.

It is a block diagram which shows the Example of this invention as a whole. It is a block diagram explaining the example of a structure of the specific expression extraction part of the above-mentioned Example. It is a figure explaining the example of the document input into the specific expression extraction part of the above-mentioned Example. It is a figure explaining the example of the morphological analysis dictionary used by the specific expression extraction part of the above-mentioned Example. It is a figure explaining the example of the morphological analysis result in the specific expression extraction part of the above-mentioned Example. It is a figure explaining the specific name extraction rule by the specific expression extraction part of the above-mentioned Example. It is a figure which simplifies and demonstrates the specific name extraction result by the specific expression extraction part of the above-mentioned Example. It is a figure explaining the example of the input document of the above-mentioned Example. It is a flowchart explaining operation | movement of the specific expression extraction part of the above-mentioned Example. It is a figure explaining the standard description table used with the specific expression extraction part of the above-mentioned example. It is a figure explaining the instruction | indication expression table used with the specific expression extraction part of the above-mentioned Example. It is a figure explaining the role table (role and particle correspondence) used by the specific expression extraction part of the above-mentioned embodiment. It is a figure explaining the specific expression record extracted by the specific expression extraction part of the above-mentioned Example. It is a flowchart explaining operation | movement of the search part of the said Example. It is a figure explaining the record searched in the above-mentioned search part. It is a figure explaining the example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a figure explaining the other example of a search by the above-mentioned search part. It is a block diagram explaining the modification of the said Example. It is a block diagram explaining the other modification of the above-mentioned Example. It is a block diagram explaining the other modification of the above-mentioned Example.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Document registration part 11 Document storage part 12 Specific expression extraction part 13 Specific expression storage part 14 Search condition input part 15 Search condition storage part 16 Search part 17 Search result storage part 18 Output part 100 Computer 121 Morphological analysis part 122 Morphological analysis dictionary memory | storage Unit 123 rule application unit 124 rule storage unit 200 client computer 300 server computer

Claims (10)

A document registration means for registering a document;
Document storage means for storing registered documents;
Search means for searching for a document stored in the document storage means based on the specific expression and the role of the specific expression in the sentence;
A document search apparatus comprising: search result output means for outputting a search result obtained by the search means.   The document search apparatus according to claim 1, wherein the search result output means highlights and presents the specific expression searched by the search means.   The search means outputs a primary search result based on the specific expression, marks only the search result that matches the role, and makes the marked search result the final search result, and the search result output means. The document search apparatus according to claim 1, wherein the document search apparatus outputs the data according to claim 1.   4. The document search apparatus according to claim 1, wherein the search means performs a search by specifying a category of the specific expression.   4. The document search apparatus according to claim 1, wherein the search means performs a search by designating a character string of the unique expression.   The document retrieval apparatus according to claim 1, 2, 3, 4, or 5, wherein the role in the sentence is a subject of an action or a state, an object of the action or a state, and others.   7. The document search apparatus according to claim 1, wherein a role in the sentence is determined by a particle and a phrase corresponding to the particle. A document means for inputting a document;
Search means for searching for a document input by the document input means based on the specific expression and the role in the sentence of the specific expression;
A document search apparatus comprising: search result output means for outputting a search result obtained by the search means. A document registration means for registering the document;
A step of storing a registered document by a document storage means;
A step of searching for a document stored in the document storage unit based on a specific expression and a role of the specific expression in a sentence;
A document search method comprising: a search result output unit outputting a search result obtained by the search unit. A document registration means for registering the document;
A step of storing a registered document by a document storage means;
A step of searching for a document stored in the document storage unit based on a specific expression and a role of the specific expression in a sentence;
A computer program for document search, wherein the search result output means is used for causing a computer to execute a step of outputting a search result obtained by the search means.

Images