JP2006031577A - Information retrieval overlooking method and device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information retrieval overlooking method and a device for dynamically calculating not only relevance to a certain keyword but also relevance between pieces of retrieved information to a retrieval target in real time, and automatically generating a new category after organizing the information to exhibit it to a user, and allowing support that confirmation of contents is performed by the relevance between the retrieved targets regardless of a matching degree to a question sentence to strengthen possibility of more efficient arrival at the intended target. <P>SOLUTION: This retrieval overlooking method for the information organizing and overlooking the information about the retrieval target on a computer comprises: a step for extracting text information from the information that is the retrieval target, and indexing it; and a step for retrieving the information matching a retrieval question imparted to input of the retrieval question from the user, calculating the matching degree to the retrieval question, making the matching degree or the relevance between the pieces of information be reflected, arranging it to a prescribed position, performing visualization, and performing overlooking display processing. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention organizes the stored information by searching for information related to the keyword specified by the user and the relevance of the searched information, and provides a search overview of the information presented to the user. It relates to a method and a device.

When searching for certain information by a computer, the information stored on the computer is searched for information that matches the specified question text, and the degree of matching (score) with the question text is determined by the VSM (Vector Space Model) method. In general, the calculation is performed by a method based on statistical probabilities such as k-neighborhood, and the list is arranged and displayed in the order of the calculated score.
In addition, for example, a search that uses information in a category in a category search such as that provided by a search site Yahoo, and displays (list display) or graphically displays the searched information based on information in the category to which it belongs. Although a method has been proposed, basically, since information in a predetermined static category is used, there is a problem that it is not possible to sufficiently respond to user requests and viewpoints.
nothing special

With the recent development of the Internet and IT technology, the generation cycle of accessible knowledge resources such as documents and databases has been shortened, and the amount of information to be accumulated is increasing explosively. However, in the situation where it is very difficult to find the intended information due to the enormous amount of information and the real-time nature that new information is added every day, it is one of the technologies to easily acquire useful information. In particular, automatic information classification technology is regarded as important.
For example, there is a method of classifying and presenting information based on definition of information relevance by hyperlinks, classification based on the relation, and category information classified in advance. However, the current situation is that static classification is not necessarily effective in acquiring information due to the above-described real-time nature of information and the difference in viewpoints of users.
On the other hand, it is desirable to adopt a method that dynamically and appropriately classifies new information, that is, a method that automatically organizes information and presents it to Just-In-Time according to user requests. It is done.

In view of the above-described conventional drawbacks, the present invention dynamically and in real time calculates not only the relevance with a certain keyword but also the relevance between the retrieved information, and organizes the information. It is another object of the present invention to provide a method and apparatus for searching and searching for information to be automatically generated and presented to a user after a new category is automatically created.
Furthermore, in conventional search systems, it is common to check the content from the one with the best score based on the score calculated by the question sentence and the degree of matching. It was very labor intensive to get there. On the other hand, in the present invention, it is possible to reach the intended target more efficiently because confirmation of the content is supported by the relationship between the retrieved objects regardless of the degree of matching with the question sentence. It is an object of the present invention to provide an information search method and apparatus for enhancing information.

The above and other objects and novel features of the present invention will become more fully apparent when the following description is read in conjunction with the accompanying drawings.
However, the drawings are for explanation only and do not limit the technical scope of the present invention.

  In order to achieve the above object, the present invention is a search overhead method for information to be searched on a computer by organizing information to be searched, and extracts text information from the information to be searched and performs indexing. Search for information that matches the given search question in response to the input of the step and the search question from the user, calculate the degree of matching with the search question, and reflect the relevance between the information and the degree of matching The information search and bird's-eye view method is configured by the steps of arranging and visualizing and displaying the bird's-eye view at predetermined positions.

  In addition, the present invention is a program for organizing information to be searched on a computer and executing a process for searching for information to be overlooked, such as PDF, Word, HTML, XML, CSV, etc. Term extraction means that automatically extracts the terms contained in the information, term classification means that calculates and classifies the relevance of the extracted terms, and information on the classified information. Relevance extraction means for classifying by detecting the feature amount and calculating the relevance of information based on the feature amount, and for the search question given via the input device in response to the input of the search question from the user Information classification means for searching for matching information and organizing the degree of matching with the search question, the relationship between the output search information and the information calculated by the classification engine and match The degree visualized by reflecting the constitute information retrieval overhead apparatus and visualization means for the processing of the bird's eye view.

  As will be apparent from the following description, the present invention has the following effects.

(1) A search overhead method for organizing information to be searched on a computer and overlooking the information, extracting text information from the information to be searched, performing indexing, and a search question from a user Searches for information that matches the search query given to the input, calculates the degree of matching with the search question, places it in a predetermined position and visualizes it, reflecting the relevance between the information and the degree of matching And a step for performing a bird's-eye view display process, so that the information search results can be viewed in an organized manner, not just in the list, so it is efficient even for a large number of search targets and search results You can get the information you intended.

(2) By the above (1), a plurality of types of information are targeted for retrieval, the relevance between the plurality of types of information is extracted longitudinally, and a bird's-eye view is obtained, thereby connecting information on the time axis. Since the hierarchical connection of information can be expressed more clearly, it is possible to efficiently understand information by knowing the connection before and after the information and the upper and lower connections.

(3) In the second to seventh aspects, the same effects as in the above (1) and (2) can be obtained.

(4) Claim 8 is also a program for organizing information to be searched on a computer and executing a process for searching for information to be overlooked, such as PDF, Word, HTML, XML, CSV, etc. Information that uses term extraction means that automatically extracts terms contained in text information, term classification means that calculates and classifies the relevance of the extracted terms, and information that uses the classified term information Relevance extraction means for classifying by detecting the feature amount of the object and calculating the relevance of the information based on the feature amount, and the search question given via the input device in response to the input of the search question from the user Information classification means that searches for information that matches the search query and organizes the degree of matching with the search question, and the relationship between the search information that is output and the information calculated by the classification means and the matching Fits visualized by reflecting, since a visual means for the processing of the overhead display, the same effects as (1) and (2) is obtained.

(5) In the ninth to eleventh aspects, the same effect as the above (1) and (2) can be obtained.

  Hereinafter, the present invention will be described in detail with reference to the best mode for carrying out the invention shown in the drawings.

In the first preferred embodiment for carrying out the first embodiment of the present invention shown in FIGS. 1 to 18, FIG. 1 shows an example of the configuration of an information terminal 1 for carrying out the information search overhead method of the present invention. The information terminal 1 includes an input device 2 for inputting data, a central processing unit (CPU) 3 for processing the input data, and the input data as it is or a central processing. A memory 4 is stored after being processed by the device 3 and an output device 5 for outputting data processed by the central processing unit 3. The information terminal 1 has the same configuration as that of the conventional one, but the input device 2, the CPU 3, the memory 4, the output device 5, and the like that are more efficient can be arbitrarily used.
The central processing unit 3 stores the initial data input via the input device 2 in a memory as it is.
Next, when a command for searching for data stored in the memory 4 is input via the input device 2, the central processing unit 3 searches for data after searching for data according to the command. Are processed and output by the output device 5.

  An example of the information search method of the present invention processed using the information terminal 1 will be described. FIG. 1 is a diagram showing a basic configuration of an information terminal that activates or processes the system, and FIG. 2 is a conceptual diagram of the configuration of the system. In the information search overview method according to the present embodiment, search of intended knowledge is performed on text information including formats such as PDF, Word, HTML, XML, and CSV, which are generally used, and as shown in FIG. Information extraction and indexing function for information retrieval and organization, and information retrieval and information organization (information relevance calculation and visualization) processing function for the purpose of structuring knowledge It consists of two processes. FIG. 4 and FIG. 5 are flowcharts showing a schematic flow of the information extraction and indexing processing function and the information organization processing function, respectively.

Here, the processing steps of the information search overhead method of the present invention will be described with reference to FIG. 6, FIG. 8, and FIG.
As shown in FIG. 6, a search overhead method of information that organizes information to be searched on the information terminal 1, visualizes and displays the information in a bird's-eye view state, such as PDF, Word, HTML, XML, CSV, etc. A term extraction step 7 for automatically extracting the terms contained in the text information to be searched through the term extraction engine 6, and the relevance of the terms extracted in the extraction step 7 through the term classification engine 8 Term classification step 9 to calculate and classify, and the feature quantity of the target information is detected using the information of the term classified in this classification step 9, and information is obtained via the relevance extraction engine 10 based on the feature quantity. Relevance extracting step 11 for calculating relevance of the user, and prompting the user (information searcher) to input a search question, searching for information matching the input search question, Information classification step 13 for classifying and organizing the degree of matching with the information classification engine 12, the search information organized in this step 13, and the relationship between the information processed in the relationship extraction step 11, Reflecting the degree of matching, each information is arranged at a position that reflects the relevance between the information and the degree of matching via the visualization engine 14. And a visualizing step 16 in which lines are arranged and organized in the present embodiment, and visualized to perform overhead view processing.
In this embodiment, the information to be searched refers to information stored in a general database, and is not particularly limited. However, depending on the place or environment in which the information search method of the present invention is used. For example, when used in universities, educational sites, etc., the information includes information for recognizing the relevance of the contents of patents, papers, etc. and their semantic similarity.

The term extraction step 7, term classification step 9, relevance extraction step 11, information classification step 13 and visualization step 16 can be basically divided into two systems for information management.
7 and 8, first, text information is extracted from an information source, ontology information is extracted from the extracted text, and indexing (indexing) is performed on the text information. Divided into processing. That is, the term extraction step 7, the term classification step 9, and the relevance extraction step 11. Then, each processing result is stored in ontology database 17 for ontology information and in index database 18 for indexing information.
Further, in the processing shown in FIG. 9 and FIG. 10, search queries such as keywords and sentences input from a user (information searcher) via an information terminal are performed by information search processing via an external search engine. The index information stored in the index database 18 is used to retrieve related information. Further, by referring to the base ontology information stored in the ontology data 17 with respect to the retrieved information, the ontology information included in each information and ontology information indicating the degree of relevance of terms are used, and information between Relevance calculation is performed, and the result is visualized on the screen. That is, the information classification step 13 and the visualization step 16.
Here, “Ontology” represents a system of concepts such as the relationship between vocabulary and vocabulary. For example, “dentist” and “dentist” are synonyms, or “dentist” is “doctor”. A logical relationship such as a subordinate concept of “,” and “dentist” defines an elemental relationship such as treating “tooth”.
“Overview” refers to observing information and its structure from a higher-level viewpoint and promoting understanding.

Hereinafter, an outline of processing work in each processing engine will be described.
The engine used in the term extraction step 7, the term classification step 9, and the relevance extraction step 11 includes the term extraction engine 6, the term classification engine 8, and the relevance extraction engine 10 as shown in FIG. It can be called an ontology extraction engine.

The term extraction engine 6 first automatically extracts a term included in text from a text extracted from text information to be searched such as PDF, Word, HTML, XML, CSV, etc. by, for example, C / NC-value method.
It should be noted that the terminology extraction engine 6 uses the knowledge of terminology in the present situation where it would be impossible to regulate 100% of the naming conventions for terms due to the vast amount of current information and the real-time nature in which they are updated daily. It is inevitable to improve the efficiency of knowledge acquisition through automatic processing. Furthermore, in situations where the definition of the term itself is ambiguous (it is difficult to distinguish between a term and a non-term), it is even more important to perform processing that involves automatic term recognition using a certain index.
Also in the system constituting the present invention, automatic term recognition based on the term recognition technology of “C / NC-value” is used as the term extraction engine 6. “C-value” is a high-accuracy automatic recognition of terms by scoring and focusing on the combination patterns of basic vocabulary related to term structure and the frequency of occurrence of terms in the target domain, and also on the nesting properties of terms. I do. In NC-value, an index of termhood is obtained using information on collocation of a candidate term with a vocabulary in the context of an actual document, and the obtained index is used as a basis. Reorder candidate terms. Our experiments show that this method can achieve an accuracy rate of 90% or higher for the top candidates for English and Japanese, regardless of domain. Such independence of the engine on the target domain and target language is an important feature for organizing and overlooking the information of the system for information in multiple languages and multiple fields.
Note that it is also possible to calculate the relevance of similar information by using the morpheme having the highest frequency or the highest frequency or higher morphemes as the vocabulary candidates for the morpheme analysis results.

The term classification engine 8 uses the term extraction result and the context information extracted from the text by the context extraction processing, and uses the similarity calculation processing between terms including the average mutual information amount to automatically classify the terms. Do.
As shown in FIG. 11, in situations where the naming convention of the term itself is unclear, the usage of hyphens for lexical conjugation and the orthographic fluctuations such as “leukemia” and “leukaemia” Variations in terms such as “NF-kappa B”, “NF-kB” acronym swings, and syntactical swings such as “human clones” and “clones of humans” Various problems arise. In order to solve such a problem, it is difficult to perform an essential solution by a method using the relative frequency of terms such as C-value in the target domain as a basic score. Furthermore, for knowledge integration, it is desirable that terms that are synonymous in meaning, such as “eye surgery” and “ophthalmological surgery”, can be automatically recognized as terms in the same class. That is, for recognition and integration of terms, automatic term class recognition technology based on a plurality of processes is indispensable as shown in FIG.
In the present invention, a technique for recognizing term variations relating to orthographic, morphological, syntactic and acronym is constructed and incorporated into the term extraction process. Further, for semantic variations, a term classification method using average mutual information is used as the statistical similarity obtained based on the co-occurrence probability with the context vocabulary.
In addition, as a method of classifying terms, not only the context information in the text of the term to be classified, but also the relationship between terms and verbs, the relationship with co-occurring terms, etc., and the vector space model etc. were used Similar term classification can be performed by using another similarity calculation method, a support vector machine, or a classification method based on machine learning such as a decision tree.

  As shown in FIG. 9, the term extraction engine 6 performs morphological analysis on text information by using a vocabulary dictionary and ontology information. Extract. Furthermore, a term score calculation process including frequency analysis is performed on the obtained term candidates. Term candidates that exceed the specified score are output as a term list.

  FIG. 12 shows a flow of processing for extracting the relationship between information based on ontology information extracted by the extraction processing. The information relevance extraction engine 10 is an engine for quantitatively calculating the semantic relevance between terms using the terms extracted by the term extraction engine 6 and the term classification engine 8 and their relevance information. Then, the semantic relevance between information is calculated by calculating the semantic relevance between terms as information characterizing the information.

The visualization engine 14 calculates the semantic relationship between information by calculating the semantic relationship between information by the relationship extraction engine 10 with respect to all the combinations between the target information. Further, when drawing the relevance of information on the screen, each information is assigned to the node N. The larger the semantic relevance, the shorter the physical distance between the nodes, and the link thickness between the nodes, that is, the association. By thickening the mark 15 and optimally arranging all target information at arbitrary locations on the screen, the relationship between the information can be visualized.
Here, the “node” is generally a junction point on a network (LAN) connecting devices such as a personal computer, a relay point on the Internet, or a computer installed there. Individual elements constituting a state in which several components are connected to each other, that is, each piece of information having a semantic relationship is called a node.

Hereinafter, as a specific example of the present embodiment, an example in which the apparatus according to the present embodiment is applied to a system for searching and searching for information such as papers and patents will be described. This system is realized by using the present invention. FIG. 13 shows a schematic diagram of the processing flow of the information search overhead system of the present invention.
The operation of the central processing unit 3 constituting the information terminal 1 according to the above-described embodiment can also be executed by a computer program described in a computer-readable language.
When the central processing unit 3 is operated by a computer program, for example, a memory for storing a program is provided in the central processing unit 3, and the computer program is stored in the memory. The central processing unit reads the computer program from the memory, and executes the operation as described above according to the computer program.
Further, by setting a storage medium storing such a computer program in the information terminal, the central processing unit reads the computer program from the storage medium, and executes the operation as described above according to the computer program. It is also possible to do so.

Next, the visualization engine 14 is used to reflect the information by reflecting the relationship between the search information organized in the information classification step 13 and the information processed in the relationship extraction step 11 and the degree of matching. In the visualization step 16 which organizes and visualizes, and performs a bird's-eye view process, a mapping that visualizes the relevance of information as shown in FIG. 14 can be displayed.
For example, when searching for “carbon nanotube” as the search information, the related text information is displayed in the window like a map, and the stronger the relationship between the information, the thicker and darker the association mark 15 is. Tied.
In addition, by performing processing as shown in FIG. 15, the relevance of information as shown in FIG. 16 is displayed for each category of “Category A”, “Category B”, and “Category C” in this embodiment. You can also.

In the present embodiment, an example of application to an information retrieval overview system for papers and patents has been described. However, the present invention is not limited to this, and for example, as shown in FIGS. 17 and 18, an existing database is used as a knowledge source. By organizing the support system in such a way that it is incorporated, it can also be utilized in a closed network such as an elementary school or a junior high school.
{Different forms for carrying out the invention}

  Next, different modes for carrying out the present invention shown in FIGS. 19 to 23 will be described. In the description of these different modes for carrying out the present invention, the same components as those in the best mode for carrying out the present invention are designated by the same reference numerals and redundant description is omitted. To do.

  The second embodiment for carrying out the present invention shown in FIG. 19 and FIG. 20 is mainly different from the first embodiment for carrying out the present invention in that real-time information is arranged in real time. In this way, re-calculation of relevance is performed, and the visualization engine 14A is used that allows the information to be viewed in just-in-time by reflecting the calculation result in the visualization. By using the visualization step 16A, it is possible to obtain the same effects as the best first embodiment for carrying out the present invention.

  The third mode for carrying out the present invention shown in FIGS. 21 and 22 is mainly different from the first mode for carrying out the present invention through the term extraction engine 6. Multilingual translation engine for extracted terms simultaneously with or after term extraction step 7 for automatically extracting terms contained in text extracted from text information to be searched such as PDF, Word, HTML, XML, CSV, etc. By using the multilingual translation step 21 which translates into a plurality of languages via 19 and accumulates in the multilingual database 20, the above-described present invention is obtained by using the information search method constructed as described above. The same effects as those of the first preferred embodiment can be obtained, and the overhead view method of the present invention can be used without hesitation in selecting a plurality of different language texts. It is possible, associations between information may also be the ones closer.

The fourth mode for carrying out the present invention shown in FIG. 23 is mainly different from the best mode for carrying out the present invention in that information to be searched is organized on the information terminal 1. And a program for executing a process for performing a bird's-eye view of information to be looked down, comprising the term extraction engine 6, the term classification engine 8, the relevance extraction engine 10, the information classification engine 12, and the visualization engine 14. Even when configured in this manner, the computer-readable storage medium 22 storing the information can obtain the same operational effects as those of the first embodiment for carrying out the present invention, and the information Even if it is used in the terminal device 1, it can be realized as a program including various commands.
In this embodiment, the term “storage medium” includes any medium that can record data. For example, the storage medium includes a disk-type storage medium such as a CD-ROM or PD, a magnetic tape, an MO, DVD-ROM, DVD-RAM, flexible disk, memory chips such as RAM and ROM, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), smart memory (registered trademark), flash memory There are rewritable card-type ROMs such as (registered trademark) and hard disks, and any other means can be used as long as they are suitable for storing programs. Can.
The storage medium 22 is created by programming each function of the microcomputer described above using a computer-readable program language and recording the program in the storage medium capable of recording the program. Can do. Further, a hard disk provided in the server can be used as the storage medium.
The storage medium 22 of this embodiment can also be created and used by reading a stored computer program with another computer via a network.
As a computer as an information terminal, a personal computer, a desktop computer, a notebook computer, a mobile computer, a laptop computer, a pocket computer, a server computer, a client computer, a workstation, a host computer, or the like can be used.

  Although the different embodiments of the present invention have been described mainly based on the first embodiment, the present invention is not limited to this, and the configurations used in the embodiments are used in combination. The same effect can be obtained.

  INDUSTRIAL APPLICABILITY The present invention is used in industries, industries, and the like that use, develop, and manufacture information retrieval methods and apparatuses.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is an explanatory diagram of an example of an information terminal according to a first embodiment for implementing the present invention. The processing flow figure of the best 1st form for implementing this invention. The conceptual diagram of the structuring of the knowledge of the best 1st form for implementing this invention. The ontology extraction process flow chart of the best first embodiment for carrying out the present invention. The search overhead view processing flowchart of the best first embodiment for carrying out the present invention. FIG. 2 is a process chart of the best first embodiment for carrying out the present invention. The conceptual diagram of the ontology extraction process of the best 1st form for implementing this invention. BRIEF DESCRIPTION OF THE DRAWINGS The schematic of the ontology extraction process of the best 1st form for implementing this invention. The conceptual diagram of the term extraction process of the best 1st form for implementing this invention. 1 is a schematic diagram of a search overhead view process of the best first embodiment for carrying out the present invention. FIG. The term classification process flow figure of the best 1st form for implementing this invention. FIG. 4 is a flowchart of a calculation process of information relevance according to the first embodiment for implementing the present invention. The schematic of the processing flow of the search bird's-eye view system of the best 1st form for implementing this invention. The reference diagram of the visualization of the information of the best 1st form for implementing this invention. The conceptual diagram of the category classification process of the best 1st form for implementing this invention. The other reference figure of the visualization of the information of the best 1st form for implementing this invention. 1 is a reference diagram of a search bird's-eye view system of the best first embodiment for carrying out the present invention. The further reference figure of the visualization of the information of the best 1st form for implementing this invention. The process figure of the 2nd form for implementing this invention. Schematic explanatory drawing of the 2nd form for implementing this invention. Process drawing of the 3rd form for carrying out the present invention. Schematic explanatory drawing of the 3rd form for implementing this invention. Schematic explanatory drawing of the 4th form for implementing this invention.

Explanation of symbols

1: information terminal, 2: input device,
3: Central processing unit, 4: Memory,
5: Output device, 6: Term extraction engine,
7: Term extraction step, 8: Term classification engine,
9: Term classification step, 10: Relevance extraction engine,
11: Relevance extraction step, 12: Information classification engine,
13: Information classification step, 14, 14A: Visualization engine,
15: Association mark 16, 16A: Visualization step,
17: Ontology database, 18: Index database,
19: Multilingual translation engine, 20: Multilingual database,
21: Multilingual translation step 22: Storage medium
N: node, W: window.

Claims (11)

A method for organizing information to be searched on a computer and searching for information to be overlooked, in which text information is extracted from information to be searched, indexed, and input of a search question from a user Search for information that matches a given search question, calculate the degree of matching with the search question, reflect the relationship between the information and the degree of matching, place it at a predetermined position, visualize it, and display an overhead view An information search overhead method comprising the steps of: A method for organizing information to be searched on a computer and searching for information to be overlooked, in which text information is extracted from information to be searched, indexed, and input of a search question from a user Search for information that matches a given search question, calculate the degree of matching with the search question, reflect the relationship between the information and the degree of matching, place it at a predetermined position, visualize it, and display an overhead view And a step of automatically extracting terms contained in the text information to be searched such as PDF, Word, HTML, XML, CSV, etc. via a term extraction engine, and terms extracted from the extraction step. The step of calculating and classifying the relevance of each other through the term classification engine and the information of the terms classified in this classification step Detecting the feature quantity of the target information through the relevance extraction engine, classifying by calculating the relevance of the information based on the feature quantity, and the input of the search question from the user, Searching for information matching the search question given through the information classification engine, organizing the degree of matching with the search question, search information output in this step, and information calculated by the classification step An information retrieval bird's-eye view method comprising: a step of visualizing via a visualization engine and performing a bird's-eye view display process reflecting the degree of relevance and matching. Visualize the relevance information of words, such as terms and search keywords, at a position that reflects the degree of relevance, and specify the related words expressed by the visualization directly with a pointing device, or automatically calculated 3. The method according to claim 1, further comprising a step of automatically creating a question sentence to be specified for a search by selecting a selected question expansion candidate from the list, and supporting a search refinement. Information retrieval overview method described. The visualization step is a step in which information can be overlooked just-in-time by recalculating relevance in real time for the organized information and reflecting the calculation result in the visualization. The information search bird's-eye view method according to claim 1, wherein: In the visualization step, each piece of information is arranged at a position that reflects the degree of association and matching between information, and a process of providing an association mark between information according to the degree of association and matching between information is performed. The information search bird's-eye view method according to any one of claims 1, 2, 3, and 4. 6. The step of calculating and classifying through a term classification engine includes the step of indexing (indexing) text information. A search overview of the information described in. The information search according to claim 1, further comprising a multilingual translation step of translating text information of information to be searched into a plurality of languages and classifying the information. Overhead method. A program for organizing information to be searched on a computer and executing a process for searching for information to be looked down on, and is included in text information to be searched such as PDF, Word, HTML, XML, CSV, etc. Using the term extraction means for automatically extracting terms, the term classification means for calculating and classifying the relationship between the extracted terms, and using the information of the classified terms, the feature amount of the target information is detected, Relevance extraction means that classifies by calculating the relevance of information based on the feature quantity, and searches for information that matches the search question given via the input device in response to the input of the search question from the user The information classification means for organizing the degree of matching with the search question, the relationship between the output search information and the information calculated by the classification means and the degree of matching are reflected. Visualized, information retrieval overhead apparatus, characterized in that it comprises a visualization unit for the processing of the bird's eye view. Visualize the relevance information of words, such as terms and search keywords, at a position that reflects the degree of relevance, and specify the related words expressed by the visualization directly with a pointing device, or automatically calculated 9. The information search overhead view according to claim 8, further comprising means for automatically generating a question sentence to be specified for search by selecting a selected question expansion candidate from the list and supporting narrowing of the search. apparatus. In the visualization means, each piece of information is arranged at a position that reflects the degree of association and matching between information, and an association mark is provided between the information depending on the degree of association and matching between information. The information search bird's-eye view device according to claim 8. A program for organizing information to be searched on a computer and executing a process for searching for information to be looked down on, and is included in text information to be searched such as PDF, Word, HTML, XML, CSV, etc. Using the term extraction engine that automatically extracts terms, the term classification engine that calculates and classifies the relevance of the extracted terms, and using the information of the classified terms, detects the features of the target information, Relevance extraction engine that classifies by calculating the relevance of information based on the feature amount, and searches for information that matches the search question given via the input device in response to the input of the search question from the user , An information classification engine for organizing the degree of matching with a search question, the relationship between search information output and information calculated by the classification engine, and matching The degree to reflect the visualized, computer-readable storage medium storing a program and a visualization engine for the processing of the bird's eye view.

Images