JP4747711B2 - Information survey device - Google Patents

Description

  The present invention relates to a technology of a search system that extracts necessary information from information such as documents and collects it.

  Many companies compete for development each year regardless of domestic and foreign countries, such as home appliances such as washing machines, air conditioners, refrigerators, digital cameras, personal computers, etc., and technologies that accompany this development. Many documents and technical information related to development have been published, and patent applications have been published. In addition, there are many technical materials such as electronic parts technology, nanotech technology, genetic engineering in the pharmaceutical industry, presentation of academic research on biological research such as brain science, publication of literature, etc. Furthermore, the range of studies that have been considered to be literary is not distinguishable from science, and related literature covers too many fields. For example, in linguistics, ranging from physics to medicine, psychology, philosophy, etc., it is extremely difficult to investigate information outside of the specialized field, not only when researching such documents on their own, but also when requesting specialists. These information retrievals are not limited to documents and patent gazettes, but are published many times on the Internet.

  In the case of patent applications for such a large number of literature information searches, the JPO provides a searchable system, and many organizations and companies also provide search systems that are generally available. Yes. Search systems include, for example, applicant search, document number search, IPC code (international patent classification) search, patent claims and full-text keyword search, and in recent years, concept searches that search based on text similarity have also appeared. ing. Some search for the inventor, or a combination thereof. On the other hand, a similar system is also provided for various presentation contents and documents, and individuals and the like can investigate from a personal computer. In the Internet world, a system that can identify where and what materials are available by keyword search is standard.

  As a conventional technique for extracting specific information from such a large amount of document information, for example, a map generation method for collecting patent information from a large number of patent information by IPC classification (see Patent Document 1), applicant and field Supporting the creation of a patent portfolio from classification of each term (see Patent Document 2), and a search system for patents and documents that allows users to select and reselect words from the search results (patents) Reference 3), synonyms are extracted and used as a database (see Patent Document 4), feature vectors are generated from document data based on the appearance frequency of keywords, and similarity is calculated to calculate the correlation between documents. What is classified and searched (see Patent Document 5) is known. In addition, the analysis based on the similarity of sentences and the analysis based on the properties of the applicant and inventor (see Patent Document 6), and the type of patent map, the inventor's correlation distribution chart is the first inventor or joint Those that are classified as inventors are known. In addition, many examples of search systems are described on the website. For example, for IVMap (trademark), it is described that a noun is extracted from a sentence such as a paper summary sentence as a keyword string, a correlation is obtained from the keyword string, and graphic display is performed.

JP 2002-175314 A (FIGS. 1 and 2) Japanese Patent Laying-Open No. 2003-76796 (FIG. 1, claim 1) JP 2002-351896 (Claim 1, 0064 column) JP 2003-108582 A (FIG. 1, column 0024) JP 2002-245067 (columns 0020, 0021, 0022) Japanese Patent Laid-Open No. 2003-296344 (FIG. 1, FIG. 4, etc.)

  As a problem to be solved, the conventional techniques for investigating information mainly consist of investigations corresponding to the purpose such as investigation of known materials. For example, in the search for patent applications, the main purpose is to examine the patent as the right holder, when it became publicly known, the content of the rights, etc., and the search to avoid or crush a specific patent The purpose of the activity was to search for documents that are similar to the text and terms of the patent from known materials prior to the filing date of the patent. However, even if we try to search the literature on the content of the rights and the full text of the patent, in the analysis such as the IPC field search and keyword search etc., the scope of the search is too wide, and the analysis is to pick up a large amount of unrelated materials and select from them There was a problem that it took too much time, a problem that the investigation was difficult, or a problem that leaked important data if the search range was narrowed down by subdividing the classification code. For example, when trying to investigate documents that use carbon dioxide as a refrigerant, there are too many irrelevant documents in the IPC field survey, while in the keyword search, carbon dioxide and symbols other than carbon dioxide are listed. There was a problem related to research techniques, such as differences in contents and symbols, numbers and alphabets, for example, if all combinations of uppercase and lowercase letters are not listed, documents that do not contain them are not included.

  When synonym information database is extracted as a synonym information database by using a search device that can specify combinations of word dependency as synonym countermeasures, the word dependency depends on the sentence, language type, or target document type, etc. There is a problem that a huge and huge database has to be constructed. In addition, when generating and analyzing feature vectors for keyword strings and sentences using nouns, keywords or keywords can be used for grasping and classifying the full text of many patent specifications, even if relatively short sentences can be easily searched. There is a problem that it is difficult to determine the degree of similarity because there are too many keyword strings, and classification accuracy cannot be obtained. Therefore, in patent search and literature search, it is common to request search skills and search specialists and search companies, and search companies also search for specific technical contents and fields for efficiency. Surveys that deviate from the intended purpose and efficiency improvement, such as being out of the conventional purpose or not being aware of survey activities in a direction that is out of focus, etc. There was a problem that the activity was excluded. In addition, there is a problem that even if you request the creation of a patent map, it will not be useful for analysis because it does not include the specific technical contents required by the user, only for abstract and general reports, such as the number of applications per year and each applicant. there were.

  Also, even if you want to thoroughly investigate a specific technology to invalidate a specific right, when you try to expand and narrow down the scope of the investigation, there is always a question of whether there is really no omission for the narrowing down, and it is almost exhaustive There was a problem that it was expensive and time consuming to continue investigating forever without being convinced. On the other hand, there is a problem for research specialists that, other than their own specialties, for example, investigators specializing in engineering technology are not confident in linguistics, psychology, medicine, etc., and it takes time and effort. Also, the person who requested the survey had a problem that taking such a survey would not only be complete, but it would not only cause dissatisfaction but also increase the risk. Furthermore, there has been a problem that it is impossible to grasp the trend of technological trends in the world simply by obtaining an enumeration of documents that disclose specific technical contents for the technical theme to be investigated. In addition, the results of investigations as required are not only that the results up to that point are terminated once the purpose of the investigation is completed, or that the results obtained by searching on the Internet cannot be checked. However, there was a problem that the survey was insufficient and the survey was broken up. In addition, there is a problem in that there is a limit in collecting information that information on the web that is announced one after another is picked up and collected only at specific time intervals. In addition, there is a problem that the search means disappears when the number of documents increases, for example, the number of pages.

  The present invention has been made to solve the above-mentioned problems. When searching for literature information, regardless of the field of science or culture, the field of investigation is enormous. It is an object to provide a device and a method that can be used for reliable and short-term investigation. It is another object of the present invention to provide an information investigation apparatus and method, and a recording medium that can grasp the trend and direction of the technical flow and record and express the contents in an easily understandable form. It is an object of the present invention to provide a reliable and practical method with good search accuracy when conducting a thorough investigation. It is another object of the present invention to provide an apparatus and method that can investigate the relationship between documents to be extracted without trouble.

The information investigation device of the present invention has completed the name and document content of the author and the inventor described in each document from each document retrieved via communication means using one or more types of retrieval formulas. A recording means for extracting and recording document information of each document including an understandable date and the document content;
From the document information of each document recorded in the recording unit, the number of documents for each specific person that is the name of the person is extracted, and a specific individual selection unit that identifies the plurality of specific individuals in the order of the number of documents,
A first module having the specific person selecting means for specifying the specific person from the literature information that is connected to the recording means and searched by a search formula including a keyword input from the input means and recorded in the recording means;
The search means connected to the recording means and searched by a search expression including a person name inputted from the input means and not including a keyword, or a search expression having a specific person specified by the first module as a person name, is stored in the recording means. A second module having the specific individual selecting means for specifying the specific individual from the recorded literature information;
Each document connected to the first module and the second module and extracted from the document information including the specific person specified by both the first module and the second module, An output summarizing means for collecting a list in the order of the number of documents and in the order of the dates;
It is equipped with .

  The present invention makes it possible to narrow down the documents easily, to easily read the results, and to perform a literature search that does not require time and effort. In addition, the present invention provides an apparatus and method capable of creating a bibliographic information map that makes it easy to understand the analysis of survey results and the flow of technological development. In addition, the present invention can efficiently perform a thorough literature search. In addition, the present invention can easily carry out a continuous survey or a continuous survey, and can make the record easier to see. In addition, the present invention can classify information at low cost and with high accuracy even for surveys accessing various information sources.

Embodiment 1 FIG.
FIG. 1 is an explanatory diagram of a system for explaining a configuration using, for example, a personal computer, which is an information research apparatus according to Embodiment 1 of the present invention. In FIG. 1, an information survey apparatus 1 is connected to a network 3 such as a LAN or the Internet, and accesses a document information server 21, a patent information server 22, or an information database 23 connected to the network 3, and retrieves document information. Or it is comprised so that the processed information may be sent out to the information investigation device 2 or the database 23. The information investigation device 1 is a control device 4 that searches information, creates a map, etc., for example, a computing unit or storage unit of a personal computer, an input means 5 that inputs data necessary for the search, a search expression or a search result to be searched The display means 6 can display the extracted list and the like, and the storage medium 7 stores the input / output results.

  The control device 4 includes a communication means 9 that is an interface for exchanging communications with the outside such as the network 3, an input / output device 8 that performs input / output processing, and the contents to be searched that are input from the input means 5, etc. There is a search expression generation means 10 that generates a search expression received via 8 and connects to an external server for communication. The literature information extracted from the information servers 21 and 22 based on the search formula and the information gathered in the other information research apparatus 2 and the information server database 23 are temporarily stored in the extraction result recording means 12. This extraction result is classified for each individual such as the author or the inventor by the specific person selecting means 11, and the date and time when the output summary means 13 creates or summarizes the presentation contents such as the document submission date, the application date, and the issue date. A list that summarizes the contents of the map and the summary contents that make the output results easy to see together with the document contents and bibliographic items in order of the publication date closest to. The collected results and information sent from other information investigation devices are recorded in the recording means 15 and can be transmitted to other information investigation devices 2 and the database 23 via the communication means 9, for example. . The output result that summarizes the literature information is sent from the input / output device 8 to the external display means 6 or the recording medium 7 such as a disk via the result display / recording control means 14 for recording and investigation in the next stage. Alternatively, it is used for review and reporting of the operator who conducts information survey.

  One of the features of FIG. 1 of the present invention is that it identifies the information of the individual who is the author of the document and the creator of the document that is the inventor of the application, and completes the content of the individual document related to the identified personal information. It is summarized in time series based on the date that the third party can understand. Articles such as papers are copyrighted, and the claimed invention has the inventor's invention rights. Like papers, patents, etc., ideas, analysis contents, technical contents, etc. are linked to individuals. Even if the right holder is pursued or the contents of the right are grasped, the individual contents as individual activities collect and analyze and examine the individual concrete contents confirmed by experiments and simulations and the process Can not grasp. In order to identify this personal activity, a literature search by a person name search that inputs the name of a specific person known from the input means 5 may be performed, or the field to be investigated and the specific content to be investigated are specified items. It may be possible to specify multiple inventors in a form that is easy to select, such as the order in which the number of references is large, a specific research institute, company, etc. Understand and summarize as a team of individuals who are identified and identified with priority. The individual with the next largest number of documents not included in this team is identified and the joint author and the joint inventor are gathered again as the next team. By separating the extracted bibliographic information belonging to each team into specific technical items, for example, those related to specific technical problems and those that are different from them, it is possible to understand the specific technical problems at a glance. It is possible to grasp trends and directions related to specific technical issues in a time-series flow of document information in which technical trends as activity results of a plurality of specific individuals are described.

  Another feature of FIG. 1 of the present invention is that an individual is first identified by a primary search, or if there are a large number of personnel to be examined, the number of individuals to be identified is limited and the result is extracted into a population that is easy to investigate. Is selected, and the entire bibliographic information of a specific individual and the technical content to be known are classified and summarized. This allows efficient general surveys first. If it is judged that the survey results are insufficient in the primary search, this particular individual, except for those who have already been surveyed, will be excluded, but the survey will be repeated several times to improve the accuracy, and the technical theme will be changed. In this way, the search for identifying the individual is performed again through a series of multi-layered surveys, such as preventing omissions in surveys and focusing on more precise technical contents. This ensures a reliable and reliable investigation. In addition, when the theme such as name and technology is first input, when the number of documents for the remaining collaborators other than those already extracted and registered is reduced, or search for already registered individuals regarding the technical theme If the display means 6 can read that the number of searches excluding the number of cases is reduced, it may be entered from the input means 5 and registered as other miscellaneous documents as it is, and the process may be terminated. Moreover, the list of output summarization results that have already been registered is read out, the contents are examined, and if the information researcher's aim is included, the end may be input.

  Further, the survey requesting side and the survey execution side have a meeting or set from the beginning to search again. This can be ensured when there are a large number of cases left or when the point you want to investigate is not included, and it can be entered manually, but it can also be done automatically without checking halfway. As an example of the case, the specified range contents to be input first are multi-layered. If the specified content of the primary search is higher in the IPC code or a broad word / phrase, enter the text as the next secondary search, or enter the specific content such as the related word / phrase, action or issue. The number of surveys can be controlled by narrowing the survey scope. On the other hand, in the case where the specific content is set as the primary search from the beginning, it is possible to take measures to prevent omission by starting from the investigation to narrow down the noise by using the high-level concept words and related words as secondary search. In this way, it is possible to collect information in a form that is easy to investigate and eliminates the waste of searching for the same specific individual many times by repeatedly narrowing down the survey in the population with limited scope and targets. . In other words, another feature of the present invention is that the name of a person is first identified by a literature survey on a specific item, and the literature information of the identified person is grouped in a time series and the specific items can be classified. Therefore, since repeated surveys exclude names that have already appeared, they are limited to the literature survey of the remaining people, and the scope of the survey is limited more and more easily.

  The input contents of this hierarchical input method are formed in a form so that they can be easily handled at the time of a survey request. If they are input as they are, they are generated by the search expression generation means 10 and the persons listed by the specific person selection means 11 However, in the next search, when the next individual is specified as it is, the number of registered individuals is excluded, so the number of search targets is narrowed down to secondary and tertiary. In addition, by narrowing down the number of people who check each stage and each level and the number of documents, the survey population for intermediate checks can be limited to a range where the literature information can be read, and this limited population check is repeated almost at the end. A high-reliability and efficient survey that not only enables quick research but also ensures thorough research because it can check the number of documents and the total number of bibliographies and bibliographic items left on the way. Is possible.

  Next, the search procedure described so far in the specific item search flow of FIG. 2 will be specifically described. First, the information search device 1 can be searched by START, for example, a personal computer is turned on and connected to the information servers 21 and 22 via the communication line 3. A specific item is input from the input means 5. An example of this specific item is described in the diagram explaining the form for requesting the survey content in FIG. The survey request form shown in Fig. 4 is an example of a document requested from the sender on the survey request side to the destination on the survey execution side or an example of a request form displayed on the screen. A column 1 item and a specific item entry column 2 for generating a search formula by the search formula generation means are described. In the column of the object for primary search in these two terms, in order to specify the search target period, for example, the survey period (b) for the last 5 years, the example of the specific range (b), and the technology that is the research theme Issue (c) is described. Depending on the information server you are accessing, you may not need to fill in the period, or it may be the full range of machine search. Or if you want to get the latest technological trends, you can use the last 3 years. Examples of specific range (b) include the designation to be the published patent gazettes of Company A and Company B as in Example 1, the investigation from individual names as in Example 3, or the stop of this range designation. good. On the other hand, the technical problem is described with a classification code, a keyword, or the like in correspondence with the technical problem as much as possible.

  Furthermore, in the example of the survey request form, the lower layer that is the second or later search that searches multiple layers is a search for the same period as the primary search, related words of the primary search technical problem, sentences for concept search, etc. Is described. Of course, upper and lower classification codes may be brought. In this example 1, the broader concept words, related situation words, and related action property words are described and expanded to the range expanded from the primary search, that is, to the vicinity of the survey range when an individual was specified by the technical problem of the primary search. There is a case where a specific range is described aiming at selecting an individual. Next, Example 2 is an example in which the same category word or subordinate concept word such as the difference in alphanumeric characters described above is described in order to investigate the missing range in the primary technical problem. Example 3 is performed in a case where the aim is off, and is performed again in order to obtain a technical problem from the beginning based on the text by the concept search. Of course, in the case where the specific person extraction is further required in the primary search, the primary search may be repeated any number of times before the secondary search. As described above, the example of a joint author or joint inventor has been described as a specific individual. However, if individual information includes personal information such as an individual name, it is obtained from the information. For example, references, books, etc. include the addition of personal names along with words of appreciation, and personal information is not limited to real names. If so, the name of a specific laboratory is also useful as the personal information of the present invention.

  First of all, in FIG. 2 (a), specific items for primary and secondary search, for example, a recent trend survey, a period of about 3 to 5 years in the past, and then the types of documents such as published patents of multiple companies are specified. Furthermore, as a technical subject, ST1 is used to input a carbon dioxide refrigerant, a supercritical refrigerant, or a keyword necessary for another investigation. A search formula is generated by the search formula generation unit 10 for the specific item input from the input unit 5, and the technical problem in the specified specific range for the last four years, for example, from the information servers 21 and 22 via the communication unit 9. Is picked up and temporarily stored in the extraction result recording means 12. The number of cases for each person name is extracted from the stored literature information, ST2, the number of person names, that is, ST3 as the first of the specific individuals, and the individual names in the order of the number of cases are specified by the specific individual selection means 11 in ST4.

  In FIG. 2, next, the joint author / joint inventor described in the literature information of the largest number of persons is identified in order ST5 and ST6. In order to identify a large number of cases, for example, about 20 people or less will be investigated first, but in order to include the joint inventor, the top 10 people in 4 years and the common top people from the top 10 people in the latest year are selected. In this way, a population that can be surveyed almost easily is obtained. That is, the maximum number of people here is about n = 10 to 20 people. It should be noted that the technical problem has been narrowed down from the beginning and that this number has not been reached due to four years, and the secondary search is continued as it is. If the search is broadened by the primary search, the search of the superordinate concept, or the search for a long period of time, only the maximum number of people set in ST4 to ST7 is repeated, and the secondary search is further narrowed down. ST8 where the specific persons extracted here are collected and taken over to ST21 in FIG. In this case, the division symbol “i” for primary search and “K” for secondary search is used, so that, for example, contents can be classified even in output summary.

  Next, as a secondary search, the primary search conditions, for example, the period of about 3 to 5 years in the past, the specification of the document type is the same, such as a patent published by multiple companies, etc. The search formula is generated again by the search formula generation means 10 with limited words such as ST9, ST10, CO2 which is the same category, and the name of a device using carbon dioxide refrigerant as a lower concept word, etc. In step ST11, the information server is accessed to extract the names and number of persons. In this hierarchical extraction configuration, the secondary or lower order may be set in advance, or the primary result may be displayed and examined, and then manually set again. At the time of extraction of the second and lower person names / numbers, a specific person name extracted in a higher-level search, for example, a primary search is excluded ST11. If the remaining number is less than a predetermined number, for example, about a few, ST14 may be output as it is ST15. In the case of these hierarchical searches, the search results in the primary search, the search results in the secondary search are collected by the output summarizing means 13 so as to include each or both, and the storage provided in the output summarizing means 13 It is recorded in the element or in the recording means 15.

  In the case of long-term search or a wide technical problem, the first search extracts a plurality of specific persons from one population, and then searches for related terms such as subordinate terms. Although the number of cases increases, ST1 to ST7 are repeated again to extract a specific person. The number of people in this hierarchy is set at K. ST12 to ST13, ST16 when inputting again as described in the figure, ST17 when searching for a tertiary or higher specified in advance is possible, and ST10 to ST17 are ST11. Repeated as the search described in. As a result, even if the setting period is 20 years long, if the technical problem is to search for the same category and narrow it down, the specific person name newly registered will be excluded from the person extracted in the previous stage. The literature can be extracted and approached zero. If technical issues are examined in this hierarchical system, almost all of the related specific persons are extracted.

  After the specific person is extracted by the specific individual selection means 11, as shown in ST 8 and FIG. 2 (b) ST 21, the documents created by the specific person by the output summarizing means 13 and stored in the extraction result recording means 12 are stored. Created as a series database. When conducting a person name survey in the secondary hierarchy, the specific person extracted from the specific person selection means 11 is specified by the search expression generation means 10 again through the input / output device 8 except for technical issues of specific items in ST1. A search is performed by designating an individual, and document information from the information servers 21 and 22 is stored in the extraction result recording means 12. However, the cases that satisfy the technical issues extracted earlier are surveys for each hierarchy, and for classification purposes, the fact that this technical issue is included is separately picked up by symbols or related words, and the literature information of that specific individual However, it is divided into whether it is related to a technical problem or not. In addition, the output result that is summarized in the output as the document information of this specific individual will be narrowed down gradually because the document information containing the specific individual described previously is excluded from the document information of the next specific individual It will be.

  This classification may be obtained by combining a search record including a technical problem such as a primary search and a specific person-created document extraction ST21, or an object of a technical problem of a primary or lower order as in ST22 for the latter. The related words may be searched again from similar word search or sentence similarity. These search and output summarizing instructions are made from the input means 5 and recorded in the input / output device 8, and are output to the output summarization means 13 via the search expression generation means 10, the specific individual selection means 11 and the result display / recording control means 14. Instructed. This selection can also be instructed from the input means 5 by designating output contents such as a patent map, and can be performed by the output summarizing means 13 via the input / output device 8 and the result display / recording control means 14. A part of the result is shown in the explanatory diagram of the list as a literature survey summary table in FIG. This bibliographic survey result summary table may be one that is sent from the sender who is the surveyer to the destination who is the survey requester, or may be a time-series database. The members are centered on the inventor D who is the most numerous and are arranged in chronological order, that is, in order of application date as a document creation date. After No. 1 to No. 8 in this list, the next inventor, for example, the inventor, such as E, F, etc., including those that have already registered D, or the next most common, such as P, Q, etc. A new name will appear as the number of inventors. In this list, the bibliographic items and contents such as the application date, the document number other than the inventor, and not only the representative figure but also the classification as to whether it is related to the technical problem are described in US. The related words may be written here. In this list, since the creators are narrowed down, duplication is eliminated and the minimum necessary document information is gathered by the output gathering means 13 (ST23).

  However, in the list of FIG. 5, the time series for each specific individual such as the inventor and the author can be seen at a glance, but it is difficult to grasp the entire time series because all the extracted specific persons are listed in order. However, if a summary of a document or a representative diagram is placed as it is in the contents column in FIG. Therefore, it is good to regroup into a multidimensional time series diagram like the patent map of FIG. In FIG. 6, the survey period is described on the horizontal axis, and the relationship between the specific person and the application described as a team is described for each central member who is the specific individual extracted on the vertical axis. Individual names may be used as long as it is difficult to divide as a team. FIG. 5 and FIG. 6 are shown as an example, and other formats, for example, a three-dimensional representation or higher representation may be used instead of a two-dimensional representation. Furthermore, after the result summarized by the output summary means 13 is displayed on the display means 6 via the control means 14, for example, the content of the document information described in the list of figures is transcribed in a blank part in the map of the figure. A device that provides easy-to-use materials by automatically entering items related to input such as survey period, keywords, etc. in the blank area of the map, and adding or deleting symbols from the input means In addition, the contents of the report summarizing the bibliographic information as the output result can be changed to the contents as desired, more accurate contents, contents desired by the survey requester, and the like.

  In the part corresponding to the vertical axis of the horizontal axis, the number of documents is listed for both the entire team or the like and the one corresponding to the technical problem. The document number may be described instead of the number of cases, and further, if the document contents of FIG. 5 can be called up and displayed on the screen, the time-series movement can be understood only by this map. The list as shown in FIG. 5 and the time series map as shown in FIG. 6 compiled by the output gathering means are recorded in the recording means 15 and can be transmitted to the outside from the communication means 9 as necessary. Furthermore, it can be displayed on the display means 6 via the result display / recording control means 14 and the input / output device 8 or can be recorded on the recording medium 7. Although the number of cases extracted from the time series map is described for each month as shown in the figure, for example, if the total number of each team for each month is described in the bottom column, it becomes easier to use.

  Fig. 2 (a) and (b) are hierarchical flows in which the same kind of classification is performed by serial survey even if there is a wide and narrow area where technical issues are searched by keywords, and a specific person in both searches is extracted. The procedure to collect and output the created bibliographic information has been explained. In FIG. 2, the keywords are primary, the personal name is secondary, the personal name is primary, the keyword is secondary, or a combination of these is tertiary or higher. Search according to various purposes and pursuit methods. Next, a procedure for outputting a combination of a plurality of different analyzes including document information extracted based on a technical problem or a person's name will be described with reference to FIGS.

  FIG. 3 is a system explanatory diagram for explaining a configuration different from FIG. 1 of the present invention. In FIG. 3, as in FIG. 1, the information investigation device 1 is connected to a network 3 such as a LAN or the Internet, and accesses the document information server 21, the patent information server 22, or the information database 23 connected to the network 3. The document information is extracted or processed information is sent to the information research apparatus 2 and the database 23. The information investigation device 1 is a control device 4 that searches information, creates a map, etc., for example, a computing unit or storage unit of a personal computer, an input means 5 that inputs data necessary for the search, a search expression or a search result to be searched The display means 6 can display the extracted list and the like, and the storage medium 7 stores the input / output results.

  The control device 4 includes a communication means 9 that is an interface for exchanging communications with the outside such as the network 3, an input / output device 8 that performs input / output processing, and the contents to be searched that are input from the input means 5, etc. There is a search expression generation means 10 that generates a search expression received via 8 and connects to an external server for communication. The literature information extracted from the information servers 21 and 22 based on the search formula and the information gathered in the other information research apparatus 2 and the information server database 23 are temporarily stored in the extraction result recording means 12. As described above, the extraction result is classified for each individual such as the author and the inventor by the specific person selecting means 11 for each document, and the output summarizing means such as the output list as shown in FIG. 9 and the inventor list as shown in FIG. It is summarized in. FIG. 10 is a summary of personal names of joint authors, joint inventors, etc. for all documents stored in the extraction result recording means 12 after being extracted by the specific person selection means 11. It is convenient to describe the list, team structure, etc., and the bibliographic items that have been searched, such as the period and search keywords. In addition, we have explained that we include joint authors obtained from literature as persons involved in creating literature as personal information, but those who were involved in exchanging information closely when creating decentralization, such as laboratories, seminars, sections, etc. If the relationship is obtained from the literature, it may be other than a person's name, so the list of the inventor list in FIG. 10 including the related person is a personal information list.

Further, in order to determine similar documents based on the contents of the full text of the document as an extraction result from the server, terms are first extracted by the term extraction means 16 in FIG. It is effective to perform linguistic analysis by morphological analysis that identifies the morpheme that is the smallest meaningful unit from the text of the document and the part of speech of the morpheme. However, particles, auxiliary verbs, etc. do not directly represent the contents of the document, and such characters can be easily removed without performing language analysis or after being analyzed as unnecessary words. Here, a noun is selected as an example of term extraction, and as a noun to be extracted, a plurality of kanji characters, concatenated katakana characters, etc. are defined as nouns, and a single kanji character and its associated hiragana or independent Concatenated hiragana and numbers are not extracted terms. Of course, a noun dictionary may be provided to select terms from this dictionary while comparing terms with this dictionary. In that case, a dictionary is provided in another microcomputer or storage device, and interaction with the dictionary is required. . The processing is convenient if a simple configuration is adopted in which unnecessary words are stored as a list without providing a special dictionary. Also, when the handling amount is large, such as full text of patent publications and collections of papers as references, the number of sentences is large and the number of nouns is enormous, and the number of occurrences of the same word appearing many times varies from 1 to many times. It will be. Therefore, when only the defined noun is extracted, the appearance number distribution of the extracted noun is represented by 0-1 value. That is, the most frequently occurring number is set to 1, and the non-appearing number is set to zero. In order to calculate a feature vector of a sentence described in a document, first, a threshold value is set for a portion of a small number of appearances close to zero, which is a small number. For example, the nouns of the appearance number intermediate part and the majority part having the appearance number distribution of 0.4 to 1.0 are extracted, and the vector generation means 17 calculates the degree of relevance of this document with other documents using this noun. The similar classification means 18 then summarizes similar literatures. The obtained display of one similar document group is to select a noun as a new keyword from a group of nouns collected from documents having a similar relevance to another document of one document, and display it in the similar term column of FIG. 9 to display or display a plurality of nouns of each similar document group as they are or by selecting a synonym in a preset method and display them as similar terms of similar documents in each document column such as the output list in FIG. Also good. As the appearance number distribution set by the threshold, the intermediate number of appearances of 0.4 was selected. This threshold is a kind of literature, and literature information covering a wide range such as general books extracts a wide range of terms and similar terms. The numerical value close to the intermediate value 0.5 is selected. In literature information such as a paper with a limited technical scope, it may be about 0.2-03, and in a patent publication in which the point of invention is further narrowed, about 0.07-0.13, for example, 0.1. The threshold value may be used. This threshold value can be selected by the input means 5 so that a user-friendly device can be obtained. However, when extracting similar documents, if the number of appearances increases, the summarization becomes easier, so the number of directions increases. It is good to set in advance with a separated middle function.

Further, a term that is excluded and not extracted by the term extraction means 16, a noun having a smaller number of appearances than the threshold of the number of occurrences, and a single Kanji character having a small number of appearances selected by using the noun threshold The dissimilar hiragana or independent concatenated hiragana, numbers, etc., and the number of appearances more than two or three are extracted as dissimilar terms by the dissimilar term classification means 19 separately. Or, when selecting a part of summaries or summaries or a combination of these without using the whole sentence, terms that have a threshold value of about 10 times or less in terms of the number of sentences may be used as dissimilar terms. The similar term classifying means 19 separately extracts the terms and outputs them in the dissimilar term column of the application list of FIG. The number of times may be set according to the amount of text in the document. That is, when analyzing similar documents or when grasping the meaning of a sentence, it is effective to use a noun string that is a term extracted by the term extraction means 16 and classified by the similarity classification means. There is a high possibility that the characteristic content of is present in a range classified as a feature vector extracted as a dissimilar term having a smaller number of occurrences. In this way, when dissimilar terms are entered in the output list of FIG. 9, terms that appear too often as already explained, such as pronouns, be verbs, articles, or foreign languages in English Paragraph numbers and the like, which are predetermined words, are excluded as unnecessary words because they do not indicate characteristics, and a noun string having a large number of appearances is also excluded from the beginning, and the one having a smaller number of appearances is selected. It is a term that is excluded from unnecessary words and similar terms and has a number of occurrences that are less than the threshold value, and is a feature term that shows a difference in similarities. In addition, a predetermined number such as at least a few times is used to emphasize the feature. Selected as the number of occurrences exceeding. In addition, after being excluded by the term extraction means 16, it is selected by the dissimilar term classification means 19 and can be selected from those not adopted as the extraction of similar terms, so that it can be realized with a simple configuration. The output summarizing means 13 includes a specific individual selecting means 11, a similar classification means 18. Individual names and terms obtained by the dissimilar term classification means 19 are regrouped into forms such as an output list in FIG. 9, a personal name list in FIG. 10, a map in FIG. 9, 10, and 11 are examples, and some of them are omitted, the contents are replaced, and further contents are added, for example, the drawings are mainly used for the document contents. Also good. In other words, in the date order closest to the date and time when the presentation contents such as the document submission date and application date were created and summarized, that is, a list that summarizes the contents and bibliographic items and the map and summary contents that make it easy to see the output results It is gathered up. In this case, the summary contents may be collected from summaries displayed in an external database, specific locations in the text, drawings, and the like.

FIG. 9 is a literature survey result summary table that summarizes a data collection created as a literature time-series database (ST21, ST38, ST53, etc. in FIGS. 2, 7, 8, etc.). It may be a detailed test data list or the like, and it may be a bibliographic time series database itself that is written on bibliographic items and collected on a time series axis for each specific individual. FIG. 10 shows an example of the result of extracting a specific person from stored document information by a specific individual selecting means. In a search module called a specific personal selection flow classified by a classification symbol i called primary search. The inventor summary table, which is a list of selected specific individuals and described as 1 searchlist at the left end, is a list of all inventors for all 132 references that have been searched for keywords in the patent publication and recorded in the extraction result recording means. The names of 350 people including the co-inventor are listed. However, although the figure shows names for 50 people here, the number of people can be freely adjusted by changing the input setting. Following this list is the name of each inventor team. That is, in the case of Mr. A's team, all the names of members who are co-inventors in Mr. A and any patent are included. In the case of the next team, all the names already listed in the previous team will be removed. The database in FIG. 9 also shows a format example described in the order of this team and in time series within the team. The format of the output contents of such a database and the like can be freely set such as a method of summarizing the team name after the time series when the amount of documents is too large. In any case, according to the present invention, related personal documents are arranged in order on a time series axis, so that the contents of the documents can be easily and efficiently analyzed. In addition, divide people into teams in order and organize them while reducing the number of people, or in other words, categorize and reduce the number of documents to be examined, and study sequentially while reducing the number of documents. It is a method that is very easy to analyze and survey the contents of the literature because it uses a population reduction method that can be analyzed. Because it can be grasped, the burden on the investigator is greatly reduced. In particular, the last is a remnant such as one per person, and it is easy to conduct a survey of all the documents that have been extracted and recorded, so even a thorough literature information survey such as an invalidation survey will only have perseverance and patience as before. It can be done easily without any problems. As described in the method of proceeding the survey in FIG. 10, the flow for selecting a specific individual from the document information retrieved by keyword as in the primary search in FIG. 2 and the first document survey in FIG. It is described as 1 searchlist. In addition, if the search expression is not a keyword, for example, in the case of the secondary search in FIG. 2 or the second search in FIG. 7, or when extracting the document information from the information server, the document is extracted and recorded instead of the keyword. In this case, 2 searchlist is described as the search flow in step 2. Of course, the number of steps 1 and 2 may be increased, in other words, the number of search modules in each search flow may be increased. For example, this search module can be classified even in the case where the document information extracted and recorded from a plurality of information servers, for example, by the same keyword retrieval formula is classified into specific individuals.

FIG. 12 is a document map similar to FIG. 6, and the time series summarization on the horizontal axis is the same as FIG. 6. On the vertical axis, a similar document group is searched by the similar classification means 18 (ST33, ST46-49), and classified by the set similar term (ST50). In addition, in the lower part where the number of cases of each similar document group is displayed in time series, the non-similar term (ST52) extracted and set by the non-extracted term classifying means 19 (ST52) indicates the feature of each document. It is described in. However, a plurality of extracted similar terms, for example, in descending order of appearance frequency may be described in the portion indicated as similar terms on the vertical axis of FIG. 12, and further, similar terms and dissimilar terms are stored in the database of FIG. Not all of the similar terms and dissimilar terms summarized in (1) may be described, but some contents may be described, or symbols may be used. In the description of the database, the document map may be made easy to see the whole picture of the document information to be investigated and analyzed, and may be compiled so that the map information can be looked at. If the necessary parts of the database can be called using the map display screen, these are technologies that can be easily performed on a personal computer such as a notation, and an easy-to-use device can be obtained. 6 and 12 are explanations for displaying the data collected by using the above-mentioned population reduction method, but when calling and displaying each similar group, all similarities to the called similar group are displayed. Make it possible to select the output format change so that it is easier to view the document. The collected results and information sent from other information investigation devices are recorded in the recording means 15 and can be transmitted to other information investigation devices 2 and the database 23 via the communication means 9, for example. . The output result that summarizes the literature information is sent from the input / output device 8 to the external display means 6 or the recording medium 7 such as a disk via the result display / recording control means 14 for recording and investigation in the next stage. Alternatively, it is used for review and reporting of the operator who conducts information survey.

  7 will be described with a specific processing flow in the apparatus of FIG. 3 as in FIG. A system in which the information research apparatus 1 of FIG. 1 can be searched by START, for example, a personal computer is turned on and connected to the information servers 21 and 22 via the communication line 3 so that specific items can be input from the input means 5. The input instruction content is transmitted to the search expression generation means 10, the communication means 9, the extraction result recording means 12 and the like via the input / output device 8, and the output summarizing means 13 via the result display / recording control means 14. Or to the recording means 15. The input specific items may be those of the primary search already described in the form for requesting the survey content as shown in FIG. 4, and FIG. There is an item entry field, and in the object field for the primary search of this item, in order to specify the target period of the search, for example, the survey period (b) for the last 3 years or 20 years, and a specific range ( (B) and the technical problem (c) that is the subject of the survey. Depending on the information server to be accessed, it may not be necessary to fill in the period, or it may be 20 years, which is the full range of machine search. Or if you want to get the latest technological trends, you can use the last 3 years. Examples of the specific range (b) may be specified to be a public gazette of Company A and Company B as in Example 1, or may be investigated from an individual's name as in Example 3, or this range specification may be stopped. . On the other hand, if the technical problem corresponding to the research purpose of item 1 is described as much as possible with the classification code or keyword, the technical problem can be easily investigated.

  Further, in the example of the survey request form in FIG. 4, the lower layer which is the secondary and subsequent searches in which multiple layers are searched is described. However, in the example of FIG. The ST38 is automatically set until the search based on the search, but the subsequent search is set by the survey request side and the survey execution side to consult based on the previous test results, or the search content is set by looking at the primary survey results. Various search routes are possible, such as not pre-entering the data. Alternatively, after identifying individuals and teams repeatedly until ST37 in FIG. 7, the second document search by the specific person extraction of ST38 is not performed, but the time series database is created as it is only for the specific person extraction, that is, the search based on the person name is omitted. It may be a flow. That is, in FIG. 7, first, a specific person related to the specified individual appearing in the document searched for the keyword and irrelevant to the already searched keyword, in other words, a specific individual including a document other than the keyword of the specific individual related to the keyword. Can be extracted and analyzed. Second, documents can be extracted and analyzed using only keywords. The first and second selections may be instructed in advance by input from the input means, or the extraction results based on the first and second flows may be stored in the database ST39.

  First, in FIG. 7, if a specific item for search, for example, a trend survey of recent technological development, is specified for the period of the past 3 to 5 years, then the type of document such as published patents of a plurality of companies to be targeted. Furthermore, ST31 which inputs the keyword of a fuel cell as a technical subject, for example. Or, if it is an invalidation search of a patent that has been granted a right, it is a period of the entire searchable range, and then, including general literature, not limited to only the published patent gazette ST31 which inputs necessary keywords such as carbon dioxide gas refrigerant, supercritical refrigerant, etc. for the content to be aimed at. A search formula is generated by the search formula generation unit for the specific item input from the input unit 5, and a document describing a technical problem in the specified specific range is picked up from the information server via the communication unit 9. The extraction result recording means 12 is temporarily stored. The number of cases for each person name is extracted from the stored document information, ST32, the first specific person is set as ST33, and the person names in the order of the largest number are specified by the specific person setting means ST34.

  In FIG. 7, next, the joint author / joint inventor described in the literature information of the largest number of persons is identified in order ST35, ST36. In order to specify the number of cases or a large number of persons, it is easy to investigate, that is, the number of cases that makes it easy to grasp the contents of the document. n is the maximum number of people in the survey, and the number of documents is limited based on this maximum value to include the persons specified in descending order and their co-inventors and authors. May be less than half. For example, we will investigate about 20 people or less first, but in order to include joint inventors, first select the top 10 people in 4 years, the top 10 people in the latest year, and select in order from common superiors etc. A surveyable population is easily obtained. In other words, the maximum number of people here should be about n = 20. If the technical problem is narrowed down from the beginning, this number may not be reached if the period is short, and the number of cases in which the person is specified from the stored number is always displayed to the display means 6. If it is displayed, it is possible to identify the total number of persons extracted by inputting the maximum number n from the input means 5 while grasping the number of remaining cases. ST34 to ST37 are repeated, and the specific person is collected and taken over to the next step.

  In FIG. 7, the first document information extracted by the retrieval specific item input and recorded in the recording means 12, that is, a specific person is first summarized in step 1 in FIG. The second-time document information, that is, the extraction in step 2 in FIG. The above processing is performed on each document recorded in the extraction result recording unit 12 by the specific individual selection unit 11 in FIG. Further, as shown in ST39 of FIG. 7, the document summarizing means 13 shown in FIG. 3 creates the document information created by the specific person including the first and second document information as a time series database and stores it in a predetermined place. . As shown in ST43, when a command for slightly changing the search formula is set for the second server search in ST38, input / output is performed based on the specific person extracted by the specific person selection means 11. The apparatus 8 and the search expression generation means 10 perform a search again in the same designated period excluding the technical problem of the specific item of ST31, and the document information from the information server is stored in the extraction result recording means 12 more than the previous investigation. It will be stored as containing literature. As described above, the flow in FIG. 7 mainly explains the processing in the specific individual selecting means 11 and the output summarizing means 13 in the configuration of FIG. FIG. 10 is a personal name list compiled by the output grouping unit 13 based on the specific individuals extracted by the specific individual selecting unit 11, and is a list of ST2 to ST7 or ST32 to ST37 in FIGS. It is a list of specific persons extracted in the range of step 1 such as the next or first search flow. ST11 and ST38 such as secondary search and second document search are within the range of step 2.

  As described above, when the search records including the technical problems performed in ST31 and ST32 are combined with the records of the search ST38 and ST39 that do not include the technical problem of extracting the specific person, the same document is excluded from the second one. The first document does not include the first document, but the number of documents examined in the second document is a little more than a few percent. For example, a symbol indicating that it has been added for the second time may be added for easy display on the document map. In the latter case, for example, in the case where document information that does not include a technical problem is added to the technical problem as in ST40, a similar word search or a sentence similarity search is performed again. . These selections can also be instructed from the input means. An example of the resulting output list is shown in the summary list in FIG. 9, the personal name list in FIG. 10, and the document map (number of cases) in FIG. This summary table of FIG. 9 is summarized by the time series axis, that is, in order of the filing date as the document creation date, by members centering on the inventor D who is the creator of the document information. In addition to the bibliographic items and contents as described above, whether or not the document is a similar document that will be described in detail later with reference to FIG. In the column of whether or not they are similar documents, similar terms are described, and the same representative similar terms are summarized with symbols such as * 1 and * 2 so that they are easy to see. Also in this list, the document creators are narrowed down in order of the largest number of items and in chronological order, and those described later are excluded from already registered document information, so the minimum necessary document information is collected by the output summarizing means 13. ST41 will be. In addition, since the document information is summarized in the document time series database, the order described in the summary table is summarized by another inventor E, or the same similar term, that is, a similar document group of symbols. Is easily performed by an input instruction. However, the time series display is the same as the newest or oldest.

  Also, the document map of FIG. 11 summarizes the number of creations in the document creator team in time series ST41. In FIG. 11, the survey period is described on the horizontal axis, and a specific person described as a team is described for each central member who is a specific individual extracted on the vertical axis. Naturally, it is possible to change the way of taking it in multiple dimensions instead of two dimensions. On the document map, symbols indicating similar document groups represented by similar terms are displayed, and dissimilar terms that are features of the documents are displayed as they are in each document corresponding to the similar document group. The display of the similar document group can be changed to another group by expanding or narrowing a range such as a threshold value by an input instruction. In addition, the team name such as the inventor and author can be changed to a team by a specified individual, and the name and order on the vertical axis can be changed accordingly. In particular, in the case of a bibliographic map, since the co-author is included, it is convenient for reading the map if the specific personal name to be called is brought to the top. These changes can be easily made based on the time series database stored in each search.

  When a similar document symbol or a feature of document content is described at a position corresponding to the number of documents or a document number at a position corresponding to the horizontal axis and vertical axis of the document map in FIG. The personal name list as shown in FIG. 10, the summary list as shown in FIG. 9, and the time series map as shown in FIG. The communication means 9 can make outgoing calls. As a result, if the survey requester is the information survey device 2 shown in FIG. 3 and the survey performer is the information survey device 1 connected via the network 3, the survey requester and the survey performer transmit and receive via the network 3. ST42 becomes possible. In both information investigation apparatuses 1 and 2, the result display / recording control means 14 and the input / output device 8 can be displayed on the display means 6 or can be recorded on the movable recording medium 7. .

  As described as the feature of the present invention, the stage before the primary search in FIG. 2 and the consultation with the survey requester in FIG. 7 can be a trial stage, as shown in FIG. 9, FIG. 10, and FIG. Use the list and map generated so that the focus can be changed in an easy-to-view form, examine the survey results, analyze and summarize the flow of technology trends, and conduct a thorough search You can connect to search.

  In FIG. 7, the survey request side and the survey execution side transmit and receive, and ST42, and by reexamining the search specific item, ST43, and a highly accurate search can be repeated again from ST31. Further, it is possible that the number of persons and the number of cases are further extracted and ST32, and the number of cases for which no specific person has been extracted may be continuously investigated as it is not necessary to review the search specific items. In addition, these operations may be set to continue automatically without the consultation ST42 by the transmission / reception of the survey requester and survey performer. That is, after generating a result such as a document map, the remaining number of surveys ST44 is performed. If there are the number of survey surveys remaining and search setting items, this information survey device 1 automatically continues the survey ST43- ST45, output and end END. All of these can be performed as a continuous operation according to the items specified in advance including the trial and subsequent continuation investigation, or the required output contents. In addition, the survey execution side who received the survey request form in FIG. 4 can independently perform trials and subsequent continuation surveys independently while judging the communication entry fields such as the survey purpose. Further, the survey requesting side may use the information surveying apparatus 2 to perform a high-accuracy search later separately from the request by only receiving the trial result. It is also possible to continue the survey for a later period only. That is, the search of the present invention can be performed easily and accurately by anyone at any time and anywhere, and a search for grasping the development progress. Furthermore, since the search can be repeated repeatedly, it is possible to automatically conduct a survey of all the cases extracted from the server.

  In this way, the present invention focuses on the activities of individuals who are continuously in charge of development, or the activities of a plurality of individuals related to the individuals, regardless of whether the search includes technical issues or not. The literature information, which is the results of the work, is summarized in time series. As a result, it is possible to change the type of search for document information, etc. related to the technical problem by combining different search flows, but not extracted by the search formula, before and after the document information classified as including the technical problem. Can be considered. In addition, as shown in FIG. 2, the search formulas for different keywords are changed, and a plurality of searches are continuously performed in series, or after the detection by the keyword as shown in FIG. 7, the search is performed again with the name of the extracted document creator. The search conditions and types can be changed, and multiple searches can be performed in series to conduct a more strict document search. For example, even if a range of similar terms can be searched for in the specific item of carbon dioxide refrigerant, it is conventional to use a term such as a higher natural refrigerant or treat it as an apparatus invention without describing the type of refrigerant. If the search system cannot be surveyed, or if the range of the classification code is expanded too much to obtain this range, the survey includes too much noise, and the number of examinations increases, which only increases the burden of period and cost. Furthermore, in searching for fuel cells, literature information on parts and materials alone, such as ion exchange membranes that do not include related terms, literature information on steam permeable membrane alone technology used in steam reforming technology, and structure of heat generation countermeasures, etc. Even if there are no terms related to fuel cells in the literature, it can be easily grasped by investigating the process of following individual development activities in time series.

  When there is a specific person to be examined in the present invention, if the person name is first set as a specific item as a trial primary, the search of FIG. 2 and FIG. In that case, a list including coauthors will be required. Also, the technical problem is set as a fuel cell in the trial, and from the literature information including the technical problem extracted by this trial, a frequently used term, for example, the word “water vapor permeable membrane” is used in one document. Alternatively, it may be automatically extracted in ST22 in FIG. 2 to be used for a plurality of documents, and the word may be added as a specific item or a specific technical problem in the next search stage to continue the search again. That is, these re-searches may be performed by a person or a method of automatically extracting words.

  Even when the survey execution side conducts the survey as a business, the search system program described in the flowcharts of FIGS. 2 and 7 may be distributed over the network 3. On the program user side, it is possible to search freely by accessing the database on the survey execution side distributed by the distributed search program. In this case, it is possible to improve the accuracy of the search by consulting with the survey expert through communication with the survey execution side such as ST42 and ST43. 2 and 7 have been described with reference to an example in which hierarchical search is performed in series, but it is possible to perform only the primary search depending on the input in FIG. Of course.

  The flowchart for explaining the operation of the information survey in FIG. 7 shows the bibliographic information and bibliographic information of the documents obtained from the information servers 21 and 22 and stored in the extraction result recording means 12 in the configuration diagram of FIG. The operation of the selection means 11 will be described. Next, the operation of processing the stored document information via the term extraction means 16 will be described with reference to the flowchart of FIG. In FIG. 8, ST31 to ST33 are those in FIG. To do. As described above, the document information extracted from the information servers 21 and 22 based on one or a plurality of types of search formulas and the information gathered in the other information research apparatus 2 and the information server database 23 are extracted result recording means 12. The term number extraction means 16 stores the appearance number distribution of the terms used in all the sentences in the document in order to determine similar documents based on the contents of the entire document as an extraction result as in ST46 of FIG. Ask. Here, a noun is selected as an example of a term, and this noun can be matched with a pre-stored noun dictionary, or the contents can be increased by learning this dictionary. Characters, connected katakana characters, connected foreign characters, etc. are defined and extracted as nouns. However, functional words representing the relationship between words before calculating the appearance distribution, that is, particles or auxiliary verbs, or words indicating content, are words that are not directly related to the characterization of the sentence. A verb list such as “Yes” and “No.” is provided with a list of unnecessary words, and these unnecessary words are removed in advance by the term extraction means 16. In the unnecessary word list provided in the storage device, particles such as te, ni, ga, wa, etc., and verbs such as having, obtained, made, etc. have been explained, but it is used to connect words and sentences. Used in relation to, for example, next, together, further, or in other words, words that connect the meaning of conjunctions and sentences, these, things, things, respectively, above, Morphological analysis to identify morphemes and morpheme parts of speech that are the smallest meaningful units from document texts, including words that emphasize the texts and sentences, especially, roughly, roughly, etc. Then, these unnecessary words are removed by referring to the list.

When providing lists or dictionaries such as noun dictionaries, synonym dictionaries, etc., select terms from this dictionary while comparing the terms with this dictionary. In that case, provide a dictionary in another microcomputer or storage device with a large storage capacity. There is still a problem that it is necessary to select a dictionary depending on which field is to be investigated. For this reason, first, nouns are selected except unnecessary words. In the case of a paper or a patent publication full sentence as a document, the number of sentences is large and the amount of nouns is enormous and the same word appears many times, so the number of appearances varies widely from one to many. Therefore, even if the number of sentences in the document is large, in order to efficiently calculate the feature vector while eliminating noise and waste, this appearance number distribution is represented by 0-1 values, and a portion having a small number of appearances close to zero, for example, a threshold value or the like. Set and do not extract. That is, it is assumed that the frequency of medium or higher is selected for similar term extraction and the threshold can be set and changed. For example, a defined noun having an appearance number distribution of 0.4 to 1.0 is extracted, and the distribution setting can be changed by the input means 5. Using the noun string, the degree of relevance of this document with other documents is obtained from the index word / document matrix. At this time, the similarity is obtained by the vector generation means 17 that performs the vector calculation ST47 weighted by the number of appearances, that is, the appearance frequency. Next, similar documents are collected by the similar classification means 18, and the display of one obtained similar document group is obtained from a group of nouns collected from a group of documents having a close relationship between one document and another document, or A noun as a new keyword is selected after the similarity that is the vector calculation result, and is displayed in the similar term column of FIG. For example, a plurality of nouns having a high frequency in each similar document group may be displayed in the respective document columns of the output list of FIG.

  When the document is expressed by a plurality of noun weight vectors and the column vector is a single document, the vector generation means 17 calculates the similarity between the column vectors, which are a plurality of documents, by using a cosine or the like. Instead of extracting the defined nouns, the keywords may be extracted by morphological analysis and the relationship between the keywords may be extracted. In this case, a vector with the dimension of the coefficient between the number of keywords and keywords is obtained for all the documents, and the feature vector is obtained from the appearance frequency in the same manner as described above, and the similarity between each document can be calculated, and clustering can be performed from the result. become. The relationship between keywords is assumed to be close to the dependency relationship obtained as a result of parsing and the distance between keywords. As an example, the distance between keywords, for example, the number of characters, the number of morphemes, the number of phrases, the number of sentences, the number of steps, the direction, etc. between the keywords is obtained, and the keywords included in the same keyword group and the relationship between the keywords are determined in the vector generation means 17. If a function for storing or learning is provided, keywords representing the category and the appearance frequency of the relationship between the keywords can be summarized. As a result, the accuracy of similarity calculation is improved and the search can be made more efficient. When the vectors of all the documents that are extracted and stored repeatedly for each document are generated for each document, the similar classification unit 18 classifies each similar document group, and the symbol similar to the symbol that represents each similar document group. Terminology is set. Although the description has been made with respect to this classification by changing the threshold value fixed by the external input, this threshold value may be set using, for example, fuzzy inference. When a specific term is extracted by removing unnecessary words after analysis of morphological elements, etc., the occurrence frequency is calculated regardless of whether the specific term is a defined noun or all other terms. The appearance number distribution is obtained. For example, specific terms having an appearance frequency equal to or higher than a predetermined threshold are used for setting similar terms for collecting similar literature groups such as patent publications, publications, and papers. As an inference engine for inferring the threshold value, it may be divided into patent document correspondence, paper correspondence, and the like. It should be noted that the engine may be divided between a similar patent search and a case where a characteristic portion which is a difference between similar patents is searched for correspondence to patent documents. The IF horizontal axis of the IF-THEN rule in inference takes the document text amount, for example, the number of document pages. The THEN horizontal axis takes a threshold value, and a threshold value that has been adopted in a past document and whose accuracy corresponding to the amount of text has been confirmed can be used for analysis of a new document. In this way, in the case of fuzzy inference, a threshold set and confirmed by an expert can be adopted rather than binary judgment, and anyone can perform the same analysis. The membership function used for inference can be set and adjusted. Depending on the type of the target document, the purpose of the survey, etc., anyone can conduct a useful survey based on the thought of a survey specialist.

  Further, in the flow of FIG. 8, the term extraction means 16 extracts similar terms, removes unnecessary words, etc., and then separates each document into a similar group without extracting the similar terms. ST51 for extracting non-extracted terms from Terms that have been removed as unnecessary words by the term extraction means 16 and have not been extracted because they are not thresholds or nouns. Nouns whose appearance distribution is smaller than the threshold, a single Kanji character and its associated hiragana, or independent Dissimilar terms such as connected hiragana, numbers, etc. that have a threshold value such that the number of occurrences is more than a very small number of occurrences such as 1, 2 and the occurrence number distribution is 0.4 or less. The term is separately extracted by the classification means 19, and this term is described in the dissimilar terms column of the application list of FIG. That is, by extracting dissimilar terms indicating the characteristic contents of each document from the similar document group, the difference between documents in the similar document group can be known. When this difference is described in the output list of FIG. 9, a term that is determined to be a noun is not the same as a term that is determined to be a noun because a term that appears too often does not show a different characteristic from that of one document. Select the one with the least number of occurrences, and the terms that are excluded from similar terms, i.e., those that are not used in similar judgments and have a low number of occurrences, and thus show differences in similarities. It is a unique feature term, and it is selected as the number of occurrences that exceeds at least twice to emphasize the feature. In addition, after the unnecessary word is removed by the term extracting means 16 and the unnecessary word is removed, it is selected by the dissimilar term classifying means 19 and can be selected from those not employed in the extraction, so that it can be realized with a simple configuration. . Furthermore, a membership function or the like may be used as a threshold for extracting this dissimilar term, and if the threshold for performing selection can be changed by external input, it can be changed according to the type of object of the document, or Anyone can efficiently conduct a highly accurate investigation by using the conditions set by the expert trying multiple times in the search. For languages with breaks such as English, it is effective to perform linguistic analysis in relation to the dictionary by morphological analysis that identifies the morpheme that is the smallest meaningful unit from the document text and the morpheme part of speech. is there.

  In the flow of FIG. 8, after a representative similar term is assigned to each document and classified into a similar document group, a specific person is extracted from each document in ST38 described in FIG. In this case, all the documents may be collected as a team by repeating ST34 to ST37 in FIG. 7, or the document creator and the document creation date may be simply summarized for each document. Next, a specific person creation document time series database is created ST53. The subsequent flow continues to ST41 in FIG. This specific person creation document time series database is created by the output summarizing means 13 and recorded in the output summarizing means 13 or the storage device in the recording means 15. For example, the output unit 11 is called from the recording unit 15 to the output summarizing unit 13 so that the order and description can be changed, and the display unit 6 or the output unit 8 is connected via the result display / recording control unit 14 or the input / output device 8. Output to the recording medium. The reference map of FIG. 11 can be created by the output summarizing means 13 from the combination of the specific person document time series data created and recorded in ST21 of FIG. 2 or ST39 of FIG. 7 and ST53 of FIG. Of course, the document map of FIG. 12 may be obtained from the specific person document time-series data created and recorded only in ST53 of FIG. 11 and 12, dissimilar terms are described for similar document groups. However, when the dissimilar terms are not classified and set, or when the non-extracted term classification means 19 is not provided. Information specific to each document is not described. FIG. 8 shows a part of FIG. 7 as a flowchart for processing in parallel. When these are combined, a series processing flow and a parallel processing flow are combined. The search flows in FIG. 7 and FIG. 8 are provided in parallel, the search in the search flow is performed in parallel based on the same survey data recorded in the extraction result recording means 12, and each time series database is created. Is done. The separate time-series databases are combined and the ST41 list is compiled as a summary table in FIG. As a database obtained by these two flows, only similar terms obtained from FIG. 8 are added to the summary table of FIG. This is because the document information of the documents recorded in the extraction result recording means is exactly the same in FIG. 7 and FIG. 8, or most of the documents are the same even if the second document survey by the specific person extraction is included, This is because the key document creation date, creator, document number, etc. are common and can be easily combined. It is not necessary to divide ST39 and ST53 separately to create a document time-series database, but it is only necessary to add data that will be obtained later. Since there may be cases, a case where search processing is performed is assumed, and a document time series database is provided for each search flow.

In this way, FIG. 7 and FIG. 8 may perform independent flow processing, or the specific individual selection means 11 which is the search flow of ST33 to ST37 of FIG. 7 and ST33 to ST49 of FIG. By connecting the means 16 and 17 for generating a term and generating a vector in parallel or in series, it is possible to easily obtain an output that is easy for anyone to understand even for a large amount of literature information. In this way, each search flow can be modularized and switched to parallel processing, serial processing, or independent processing, by changing the connection of the device, by automatic circuit selection from input settings, or automatically from the required output A convenient device capable of executing the search processing can be obtained by selecting the circuit correctly or by connecting all the search processing circuits without switching and always operating them. Moreover, by providing a document time-series database ST21, ST39, ST53 for storing the processing results obtained for each search flow, or by providing a basic database that describes the document creation date and time, creator, document number, etc. By additionally describing the data from each search flow, the database for obtaining the output can be easily synthesized. In creating the document time-series database, it is possible to divide each, or just add data obtained later.

  In other words, the flowchart of FIG. 2 and the flowcharts of FIG. 7 and FIG. 8 have described the flowcharts in which a plurality of search flows are combined in series or in parallel. However, each search flow is processed and the results are collected. A structure may be provided in which information survey devices 1, 2 and the like that individually hold each module independently are provided. Alternatively, the search flow of FIG. 1 or FIG. 3 is connected in series or in parallel to the information investigation device 1 and the flow of FIG. 2, FIG. 7 and FIG. It is also possible to select a process by the input means 5 or to conduct a search by each search flow which is a plurality of modules. Or it is good also as a flow which combined FIG. As shown in FIGS. 2, 7, and 8, in which different search processes are performed in different configurations of the present invention, the flows as shown in FIGS. 2, 7, and 8 are integrated into each individual flow or partial flow connected in series, or a tertiary or higher search flow. For example, as shown in FIG. 13, it is possible to combine modules, which are each search flow, into a search network in a single mesh pattern. FIG. 13 shows an example of a search network structure shown as an example. This search network may be formed in a single personal computer, or may be configured by communication with a plurality of personal computers. As described above, a network circuit is formed by combining a plurality of search flows such as the same process and different processes for searching document information as modules. 13, 14, and 15, the search process contents and the search ranges of information to be searched are the same or different, and the primary keyword search, the primary person name search, the secondary keyword search, the secondary person name search, A search flow module such as similar document group search or dissimilar term search 103-108 is used as a node of the network circuit, and it is an output side node from the search specific item input means 101 which is an input side node and the literature information acquisition means 102 from the server. Connected to the time series database creating means 109 and the output means 110 to form a network and a work circuit.

  FIG. 13 to FIG. 15 are diagrams showing a method of processing document information and search flows having the same or different contents connected via a network. That is, a primary keyword search, a secondary keyword search, etc. that are search flows (ST3-7, ST8, ST11, etc. in FIG. 2) when a specific person is extracted from the document information extracted and recorded by keyword search and made into a database. There is a second search module. It should be noted that a primary person name search, a secondary person name search, etc., which are search flows (ST3-7, ST8, ST11, etc. in FIG. 2) when extracting a specific person from document information extracted and recorded by person name search and creating a database, etc. There is also a search module that has the same flow as the previous one in terms of the processing procedure of selecting a specific person, but has different reference information to be analyzed that is the starting point 102. Further, the similar document group search flow and dissimilar term search flow (ST33, ST46-ST50, ST51, ST52, ST38, etc. in FIG. 8) by the configuration of the term extraction means shown in FIG. There is a second search module having the following procedure. Note that the dissimilar term search flow may be the third search module, or a network in which different fourth and fifth search modules are provided by changing the language configuration analysis method of the similar document group search flow. That is, in the information processing device, the information processing means extracts a plurality of different search modules, and further, these modules are further divided into subdivisions and extracted by the keyword and the module 1a extracted by keywords for obtaining the literature information at the starting point described above. Module 1d for classifying module 1b, or classifying module 1c extracted and recorded from one information server with a keyword different from module 1a, or classifying one extracted and recorded from a plurality of information servers with the same keyword It may be. In this way, a network structure for performing search processing is formed by processing document information by combining modules having different search processing procedures. Further, even if the search processing procedure is the same, another network-like network structure is formed by providing a plurality of modules for changing the document information to be processed. Further, both such networks may be combined. Next, such a search network structure will be described.

13 to 15, each search flow has the same or different search structure and search range, and search flow modules such as 103-108 are used as network nodes. The network circuit is connected from the means 102 to the time-series database creating means 109 and the output means 110 which are output side nodes, and flows information. Search specific item input means 101, 111, 121 as starting points in FIGS. 13 to 15 perform processing of the input / output device 8 and the search expression generation means 10 in FIGS. Further, the server which is one end of the net of the net structure performs the processing in the communication means 9 and the extraction result recording means 12 in FIGS. 1 and 3 with the document information obtaining means 102, 112 and 122, ST2 in FIG. This includes document extraction operations such as ST11 and ST32 and ST38 in FIG. The time series database creation means 109, 119, and 129 that are the other end of the network of the search net structure are connected to the output processing means 110, 120, and 130 that are provided in the output summarization means 13 and serve as exits. ST39 of FIG. 8 and ST53 of FIG. Each search flow module that forms a contact point of a mesh in FIG. 13, that is, a node and performs a specific search process, is a search flow that is an individual module described with reference to FIGS. Different types of search modules 103, 104, 105, 106 different from 102 are connected. Each search flow module of the primary keyword search 103, the primary person name search 104, and the similar document group search 105 that can be the main search can receive the document information from the starting point 102 of the mesh and perform the processing in each search flow in parallel. The search contents can be recorded in the time series database 109 as shown in FIG. 9 or the search flow results can be recorded together. Each secondary search flow in FIG. 13 is connected in series with the main search flow, and all search flows are connected to the creation of a time-series database that is the other end of the net of the net structure. In FIG. 13, only the primary is described as the similar document group search flow. In the network structure shown in FIG. 13, complicated processing such as primary and secondary can be easily performed while changing the investigation range.

The connection of each search flow module shown in FIG. 13 is changed, and it is possible to reciprocate between the starting point 112 and the main flows 113, 114, 115, etc. as shown in FIG. You can retake and perform other types of searches. On the other hand, only the dissimilar term search flow 116 is a one-way information flow with respect to the starting point 112, but a route from the similar document group flow 115 exists. Note that FIG. 14 forms a network structure in which document information is collected by a module having only a primary search flow. Further, in the network structure of FIG. 15, the primary search and the secondary search, which are the same type of search, are connected in parallel without being connected in series to increase the processing capability. FIG. 13 is a search net structure in which the first module is provided with a plurality of series circuits in parallel, and information flows in one direction from the entrance to the exit, and a configuration that can easily cope with any information. Show. In FIG. 14, a plurality of first modules, second modules, etc. are provided in parallel, and the flow of information to and from the entrance is bi-directional. A simple structure is possible by simply selecting whether or not to return. Of course, the network structure of FIG. The thing of FIG. 15 has shown the structure which can output various with respect to a large amount of information.

  In the search network structure shown in FIG. 13 and the like, literature information downloaded and stored from the net 3 at the input side, particularly at the starting points 102, 112, and 122, is supplied to the network structure that is a mesh, and the information is processed in each module. A database is created on the output side 109, 119, 129. This network structure is a structure in which each module node is provided almost regularly from the input to the output. The keyword search flow, the name search flow, and the similar document group search flow are not only primary but secondary, and higher order. Is also possible. However, nodes that connect flows of dissimilar term search are irregularly provided, and it is also shown that a structure in which the information flow direction is an irregular flow is included. FIGS. 13 to 15 describe the processing contents of processing the document information by combining modules having different search processing procedures. For example, each search module in FIG. 13 is a keyword search that is a process based on personal information. Also good. In that case, there will be no irregularities such as nodes connecting dissimilar term search flows, but the number of nodes passing through to the end point, which is the exit described in FIG. 13, is different, that is, a regular network is not used. If a circuit having a long path and a short path is provided, the content of information processing is increased and the transmission efficiency is improved. However, if the number of long circuits is reduced, the configuration becomes simple. In this way, a partially unbalanced network structure having each search flow as a node is incorporated in the apparatus that conducts the information survey, and as described above, the document information is flowed to all. However, the configuration may be such that each search flow can be selected or disconnected by control of the control device 4 by setting the input means 5 or can be freely flowed without controlling the flow of the document information. In addition, by combining the flowcharts as shown in FIGS. 2, 7, and 8 such as enlarging or reducing the mesh structure, or by making each independent, or by processing a plurality of paths simultaneously. A search network structure is created that allows not only different surveys, such as different ranges of surveys, different content surveys, different purpose surveys, but also different results or multiple different results. In addition, at least the time series databases 109, 119, and 129 that are the exits of the network include individual search results that are the results of performing the search process simultaneously with all the nodes of this search network connected, or all search results. Since the minimum necessary information is recorded in a collected state, various output contents can be output and displayed. In this case, there is a common code for the document creation date and time, the creator, the document number, and the like, and they are easily summarized. Alternatively, the search-processed data from the nodes in each search flow is collected in the time-series database creation 109, 119, and 129, and a common part and a part added at each node are output from the collected data. It is also possible to adopt a configuration in which the data is output in the required output form or standard output form.

For example, in this network structure, efficient search to obtain the target information by route selection, which is a search flow such as all nodes and partial switching in this network structure, thorough investigation to obtain documents describing the target information, wide problems and specific problems Or a trend survey such as obtaining a flow of past information of a specific individual, etc., and a device that enables many different searches. Such passage selection can be anything such as specifying the passage with the input means 5, specifying the output contents, classifying the type of input, or setting the standard output to be preset without specifying anything. is there. The literature information flows through the search network according to these designations and classifications, and is processed and output. In FIG. 13, the primary and secondary hierarchies are used. However, this may be a multi-hierarchy structure of the third or higher order, or only the primary parallel processing structure or any one search module as shown in FIG. A simple network such as one in parallel with one other search module may be used. Further, a plurality of pieces of information stored at the starting point 102 can be provided with respect to FIG. In this case, different document information of the same retrieval formula obtained from a plurality of information servers 21 and 22 is classified, and the information retrieval processing is automatically classified into the primary and secondary, or the flow is vacant with a time difference. Flexible processing such as flowing and processing individually becomes possible. By taking this configuration, the investigation device of the present invention automatically obtains information while connected to the Internet, etc., and repeats the procedure of processing in a different module each time information is obtained. By displaying the information on the information time-series map as shown in FIG. 11 or FIG. The continuous information survey can be automatically performed by such continuous operation.

Furthermore, although the similar document group search 105 is described as being provided in parallel with each primary search flow for server document information acquisition, it may be provided before and after each primary search flow to have a configuration such as serial connection. Further, although not shown in the figure, a search network structure formed by a node configuration in which the presence / absence of a transmission / reception path for consultation such as ST42 shown in FIG. With the configuration in FIG. 13, for example, even for the technical problem with the same content in FIG. 2, a plurality of hierarchies such as wide and narrow keyword settings can be investigated, or a survey of only a single hierarchy can be selected. As shown in FIG. 7, it is possible to select, for example, a survey of technical issues and names, and a survey by extracting similar terms. A combination of FIG. 2 and FIG. 7 is also possible, and keywords and personal names can be searched in multiple layers. However, the dissimilar term search 106 is used in combination with the similar document group search 105. In addition, search using each node is performed in various states, and the time series database created by a specific person is recorded on a recording medium, and this recording medium is carried around and varies in various places or at multiple survey members. It is also possible to conduct surveys in parallel at the same time. In the search network structure shown in FIG. 13, the structure that can reciprocate between the starting point 102 and the nodes 103, 104, etc. has been described. Also good. These can be formed as long as they can be formed in the flowcharts of FIGS. 2, 7, and 8.

In FIGS. 13 to 15, each means and the search flow module number are changed for easy understanding. However, FIGS. 13, 14, and 15 show a state in which a passage through which document information flows is switched. That is, it shows a state in which the mesh connection is changed, and for example, a structure in which many modules are provided may be changed to the mesh state in this way depending on the output contents. In the example of FIG. 13, the flow 103 only, the flow 104 only, the flow 107 or 108 via the flow 103 in the example of FIG. 13, the flow 113 via 114 in the example of FIG. 14, or the flow 123 in the example of FIG. There are many selection patterns such as 127 via 128 and 124. Further, for the document map of FIG. 12, in the example of FIG. 13, in the example of FIG. 13, from 107 to 108 via the flow 105, in the example of FIG. 14, from 115 to 116 via the flow 113, or from 115 to 116 via 115, etc. Flows 125, 126, and 127, or 125, 126, and 128 can be selected. The selection of these patterns may be switched by setting the input means 5, or such a pattern diagram and the output contents of the pattern can be displayed on the screen to select which pattern is desired. Each pattern may be operated in parallel according to the flow of information and processed to create a time-series database, which may be combined on the time-series database. Furthermore, it is also possible to prioritize the shortest processing such as stopping other processing when a database is obtained by operating each pattern in parallel and creating a time series database to obtain the requested output contents. . Alternatively, it may be possible to record all the time series databases in each processing and cope with a change in the output content request. Many patterns can be formed for the output FIG. 11 such as the flow 105 via the flow 103 or the reverse flow, or for the output FIG. 12 such as the flow 106 via the flow 105. As a feature of the network structure in the present invention, a time series database is financed to the search module which is each flow, and the data gathered in each search module is added to the next module and data not found in the previous one, and the time series of the end point of the exit Since all information is aggregated in the database as shown in FIG. 9, the final output can be quickly obtained in the form of the request even with the processing result by the free information flow.

  As a network structure with a simpler configuration, referring to FIG. 14, the primary keyword search module 113 is in one direction between the start point 112 and the end point 119, and the same 113 is arranged in parallel, or the start point And 113 and 114 are arranged in parallel between the end points. Alternatively, it can be said that the structure between the start point and the end point is only the structure 115 and 116 with a part omitted in FIG. FIG. 16 shows a schematic configuration diagram in the case of such a configuration. FIG. 16 is an explanatory diagram of a system for explaining a configuration different from those of FIGS. 1 and 3 according to the present invention. In FIG. 16, as in FIG. 3, the information research apparatus 1 is connected to a network 3 such as a LAN or the Internet, and accesses the document information server 21, the patent information server 22 or the information database 23 connected to the network 3. The document information is extracted or processed information is sent to the information research apparatus 2 and the database 23. The information investigation device 1 is a control device 4 that searches information, creates a map, etc., for example, a computing unit or storage unit of a personal computer, an input means 5 that inputs data necessary for the search, a search expression or a search result to be searched The display means 6 can display the extracted list and the like, and the storage medium 7 stores the input / output results.

  The control device 4 includes a communication means 9 that is an interface for exchanging communications with the outside such as the network 3, an input / output device 8 that performs input / output processing, and the contents to be searched that are input from the input means 5, etc. There is a search expression generation means 10 that generates a search expression received via 8 and connects to an external server for communication. The literature information extracted from the information servers 21 and 22 based on the search formula and the information gathered in the other information research apparatus 2 and the information server database 23 are temporarily stored in the extraction result recording means 12. In FIG. 16, it is described that the specific individual selection means 11 shown in FIG. 3 is removed. The full text of the document as the extraction result from the server is sequentially extracted from the extraction result recording means 12, and the term is first extracted by the term extraction means 16 in FIG. Linguistic analysis is performed by morpheme analysis that identifies the morpheme, which is the smallest meaningful unit from the text in the literature, and identifies the part of speech of the morpheme. First, unnecessary words are removed. For example, particles and auxiliary verbs do not directly represent the content of sentences, and such characters can be easily analyzed by removing them as unnecessary words without performing language analysis. Here, a noun is selected as an example of a word extracted by the term extraction means 16, and a plurality of kanji characters, concatenated katakana characters, etc. are defined as the noun to be extracted, and a single kanji character and its connected hiragana are defined. Alternatively, independent connected hiragana and numbers are not extracted terms. Of course, a noun dictionary may be provided to select terms from this dictionary while comparing terms with this dictionary. In that case, a dictionary is provided in another microcomputer or storage device, and interaction with the dictionary is required. . The processing is convenient if a simple configuration is adopted in which only unnecessary words are stored as a list in a storage device such as a microcomputer without providing a special dictionary. In addition, when the handling amount of patent publications such as full texts of patent publications and collections of papers is large, the number of sentences and the amount of nouns are enormous and the same words appear many times. The number of appearances varies widely from one to many times. Therefore, when only the defined nouns are extracted excluding unnecessary words, the distribution of the number of appearances of the extracted nouns is represented by a 0-1 value. That is, the most frequently occurring number is set to 1, and the non-appearing number is set to zero. In order to calculate a feature vector of a sentence described in each document in order to investigate similar documents from the extracted noun string, first, nouns with a small number of appearances are removed by a threshold. For example, a noun lacking in many parts is extracted from the middle part of the number of appearances having an appearance number distribution of 0.4 to 1.0, and the degree of relevance of this document with other documents is transmitted to the vector generation means 17 using this noun. The vector is calculated, and the similar classification means 18 recompiles the similar literatures. The display of the obtained one similar document group may include selecting a noun as a new keyword from the noun group collected from documents having a close relevance to one document. After the unnecessary words are removed by the term extracting means 16, non-extracted terms can be further extracted from the unextracted terms by the non-extracted term extracting means 29 and collected as feature words of this document. In the case of patent documents such as published gazettes with a small amount of text, the extraction of non-extracted terms is not performed with no threshold, or the extraction of non-extracted terms is not performed with the threshold set to multiple times with a small number of digits. Analysis is possible and effective by simply classifying.

  The specific processing content of the document information processed by the information processing apparatus configuration of FIG. 16 will be described with reference to the flowchart of FIG. In FIG. 16, bibliographic items for each document, which are the keys necessary for database creation, can be obtained in the same manner as in the prior art, up to ST31 to ST33 in which the retrieval specific item is input, the number of persons is extracted, and the document number i is set. The document stored in the extraction result recording unit 12 is subjected to morphological element analysis of the document in order to calculate the similarity of the sentence by the term extraction unit 16, the vector generation unit 17, and the similarity classification unit 18, ST61, and unnecessary words are removed ST62. In order to obtain the number of used nouns defined from all sentences in the document, ST46 calculates the noun appearance number distribution. The noun can be obtained by associating with a noun dictionary stored in advance, or the contents can be increased by learning the dictionary. Here, for simplification, multiple kanji characters that are not one character, connected katakana characters, connected foreign characters, etc. are defined as nouns and extracted. However, before calculating the distribution of the number of occurrences, a list of unnecessary words should be removed beforehand by providing a list of unnecessary words that are not directly related to characterizing the text even if they are functional words that indicate the relationship between words or words that indicate content. Keep it. As an unnecessary word list, it is used to connect verbs, words, sentences, etc. In this regard, for example, next, together, the words that connect the meanings of the conjunctions and sentences, such as, or i.e., these, things, and things, respectively, emphasize the above, the words, and the sentences. It includes words, in particular, almost, roughly, etc., and performs a morphological analysis that identifies the morpheme that is the smallest meaningful unit from the text of the document and identifies the part of speech of the morpheme, and lists these unnecessary words And will be removed.

  In the case of a paper or patent publication full text as a document, the number of sentences is large and the amount of nouns is enormous and the same word appears many times, so the number of appearances varies widely from one to many. In order to efficiently calculate feature vectors without noise and waste even if the number of texts in the document is large, this appearance number distribution is represented by 0-1 values, and a portion with a small number of appearances close to zero is set, for example, as a threshold ST63 that determines the appearance rate such as not extracting. If this appearance rate is large, for example, a medium or higher frequency is selected for similar term extraction, and a noun or a noun that is not defined as a low appearance rate is extracted separately as a non-extraction term ST64. It is assumed that a simple threshold is set for the appearance rate determination ST63 and that the threshold can be freely changed. For example, a defined noun having an appearance number distribution of 0.4 to 1.0 or more is extracted, and using this group of noun strings extracted for one document, Is obtained from the index word / document matrix. At this time, the document text vector calculation ST47 weighted by the appearance frequency, that is, the appearance frequency is performed to obtain the similarity between documents. When similarities are obtained for all documents and similar documents are collected, similar terms are set for display of the obtained similar document group ST70. As a similar word, a noun that is a new keyword may be selected from a group of nouns collected from a group of documents having a similar degree of relevance between one document and another document, or a similar word that is a vector operation result, Extracted and intense terms with a higher appearance rate may be used as they are. In addition, non-extracted words such as extracted noun strings with a low appearance rate are also set, and a term database in a similar format that has the same items as ST65 and the specific person creation time series database creation ST53 of FIG. It is summarized as ST66. For example, FIG. 18 and FIG. 19 show examples in which a plurality of nouns having a high frequency in each similar document group are collected as representative similar terms of similar documents as they are, and terms in one document are collected. In addition, although the appearance number intermediate number 0.4 was selected as the appearance number distribution set by the threshold value, this threshold value extracts a wide range of terms for document information covering a wide range such as a general book. Therefore, it is suitable to combine similar terms, and a value close to the intermediate value 0.5 is selected. In literature information such as a paper with a limited technical scope, it may be about 0.1-03, and in a patent publication in which the point of the invention is further narrowed, about 0.07-0.13, for example, 0.1. The threshold value may be used. Furthermore, as will be described later, a small threshold is desirable even when the contents of one document information are collected. The judgment of the appearance rate may be a threshold value, but fuzzy inference may be used because it is a 0-1 value. In that case, if the numerical value can be selected by the input means 5, it is easy to use.

Also, it is a term other than an unnecessary word that has not been extracted by the term extraction means 16, for example, a noun that is defined as the number of appearances that is less than the threshold of the appearance number distribution, and a verb that is excluded outside the noun definition range. A single kanji character and its concatenated hiragana, or independent concatenated hiragana, numbers, etc. These are extracted separately by the non-extracted term extraction means 29 on the condition that the number of appearances is more than a few. That is, when analyzing similar documents or when grasping the meaning of a sentence, it is effective to use a noun string that is a term extracted by the term extraction means 16 and classified by the similarity classification means. There is a high possibility that the characteristic content of is present in a range classified as a feature vector extracted as a dissimilar term having a smaller number of occurrences. Therefore, it is preferable to select a verb that has not been extracted as outside the noun definition range and that has a smaller number of appearances in relation to the threshold of the explanatory drawing. When extracting non-extracted terms in this way, terms that appear too often as already explained, such as words that have been pre-determined according to the sentence format in English regardless of pronouns, be verbs, articles, or Japanese foreign languages The above paragraph numbers and the like are excluded as unnecessary words because they do not indicate characteristics, and a noun string having a large number of appearances is also removed from the beginning, and the one having a smaller number of appearances is selected. It is a term that is excluded from unnecessary words and similar terms and has a smaller number of occurrences than the threshold, so that it can be treated as a feature term that shows a difference among similarities that are represented by similar terms that have already been extracted. In addition, the number of appearances exceeding a predetermined number, such as at least two or three times, is selected to emphasize features. Since it is selected by the non-extracted term extracting means 19 after being excluded by the term extracting means 16, it can be selected from those not adopted at the time of extraction as similar terms, so that it can be realized with a simple configuration. In the output summarizing means 13, a database is created and ST66, and already-explained document maps and others are created ST67. In the database, the personal names and terms obtained from the literature information are regrouped into forms such as an output list in FIG. 9, a personal name list in FIG. 10, a map in FIG. 9, 10, and 11 are examples, and some of them are omitted, the contents are replaced, and further contents are added, for example, the drawings are mainly used for the document contents. Also good. In other words, in the date order closest to the date and time when the presentation contents such as the document submission date and application date were created and summarized, that is, a list that summarizes the contents and bibliographic items and the map and summary contents that make it easy to see the output results It is gathered up. In this case, the summary contents may be collected from summaries displayed in an external database, specific locations in the text, drawings, and the like.

FIGS. 18 and 19 are diagrams for explaining typical examples of the database created by the apparatus configuration of FIG. 16. For simplicity, the extracted terms in a state in which no vector calculation for obtaining similar terms is used are used as they are. ing. FIG. 18 shows an example in the case of processing a plurality of documents. FIG. 19 shows an example of a database that can introduce book contents with similar terms and non-extracted terms for each province, which are sentence delimiters. Explain that it is true. FIG. 18 shows the term search results as a database, and there are bibliographic items such as the document creation date and time and the document number as in the document search result summary table of FIG. 9, and there is no description of the creator who is a specific individual. It can be compiled into a common database along with other types of search results. In the contents column, extracted terms that are nouns with a high appearance rate and unextracted terms extracted from outside the extraction range are described, but as already explained, these terms are classified into similar terms and similar terms. It can also be expressed as Further, this document is a patent publication, and a representative diagram thereof is also described. The most frequently appearing words / phrases in the document information 1 are keywords, and those having an appearance frequency threshold of 0.2 or more are sequentially extracted term noun strings, and the contents of this document can be roughly grasped. . Non-extracted terms selected from outside the range of extracted terms were selected with a threshold of 0.1 or less and an appearance frequency of 6 times or more, but the content to be calculated and processed was understandable. In the end, it is understood from this database that it is possible to grasp that the feature vector is calculated based on the appearance frequency of the relationship between the keywords / keywords, which is desired to be asserted as this document if the phrase analysis and similar classification are performed on this document. Even document information based on a chart can use the system of the present invention based on terms in the chart, and a chart based on similar terms may be described as document contents.

FIG. 19 is a term database similar to FIG. 18, but is a map of term search results for investigating the contents of a book fuel cell. Reference information in terms of extracted terms and non-extracted terms for each chapter that divides the contents of a book, not in time series. Is a survey. The threshold for obtaining the extracted term noun is about 0.1-0.2, and the number of appearances smaller than this threshold is selected to obtain the non-extracted term. Chapter 1 shows the number of occurrences of hydrogen. From this map, it can be seen that Chapter 1 contains a general description of the fuel cell and further characteristic relationships. Chapter 2 shows that patents, ballots, and fuel cells have almost the same number of appearances, and that there are not so many distribution states, and there are explanations related to patents and automobile companies in fuel cell development. It can be seen that Chapter 3 explains the specific relationship between the car and the fuel cell because the number of appearances of the car is large. In the fourth, the number of occurrences of hydrogen has been skipped, but it can be seen from other extracted terms and non-extracted terms that it describes the hydrogen infrastructure for use in fuel cell manufacturing and automobiles. By analyzing sentences related to terms such as hydrogen, which is divided into methods and infrastructure, natural gas as fuel and gas generated during production, energy used in various explanations, etc. It was found that the chapter contents can be grasped more accurately. In other words, not only can the similar function and the non-extracted terms be summarized for each sentence of the electronic book, but the table of contents function of the book can be fulfilled. On the other hand, if it is configured to call each chapter, it can be used by anyone anywhere as an electronic book-like device with a search function that can read the desired location. In other words, the information survey apparatus of the present invention is not only used for time series summarization, but the vertical axis not only searches for similar document groups by the similar classification means 18, but also uses the classification of similar terms and non-extracted terms. It is also possible to grasp the contents of the more accurately. In addition, in order to see the similarity of the whole sentence, terms that appear frequently are extracted and similar terms are obtained, while when grasping characteristic contents from similar documents and similar sentences, it is possible to judge from the similar words as a whole, It was found that there is a method to extract and judge terms that appear less frequently. Therefore, it is effective to grasp the contents of the document information using a map in which both similar terms and non-extracted terms are described. It should be noted that similar terms and non-extracted terms collected in the database may not be described in all words, but may be a partial description or may use symbols. In the description of the database, the document map may be made easy to see the whole picture of the document information to be investigated and analyzed, and may be compiled so that the map information can be looked at.

  As described above, the present invention has a simple configuration of a network composed of a single type of search module, or a combined network structure of modules including different search processes. It is a structure that can be collected in time series, extracting specific terms from appearance frequency, obtaining similar terms and extracting non-extracted terms from outside this extraction range, etc. It is useful for surveying a vast amount of information such as easy to read and easy to understand, and thorough investigation. Furthermore, it is possible to investigate from a plurality of information sources at the same time, or to collect a method and information that have not been considered before, such as automatic continuous survey. Further, the similar term search or similar document group search flow 115 and the dissimilar term search or dissimilar term extraction flow 116 are connected in one direction between the acquisition of the document information as the start point 112 and the creation of the database as the end point 119 in FIG. In other words, another use is possible by the simplified network structure from the start point to the flow 115, from 115 to 116 and the end point 119, and from the last 116 to the end point. That is, for complex legal documents such as a single contract, or a large number of books, instruction manuals, etc., but there is only one document information that cannot be easily read. As an output via this network of the present invention, it can be replaced by the number of pages instead of time series, the number of chapters, items, the number of contract clauses, etc. for dividing sentences. As the output, the display content equivalent to the document map of FIG. 19 is obtained, such as the number of pages on the horizontal axis, similar words in sentences on the vertical axis, and dissimilar terms described for each page or article. That is, the contents of one document in this format are displayed in a form that is easy to understand or analyze. As a result, since such an output can be used as a table of contents for analyzing complicated contracts in electronic commerce and books using electronic information, analysis in a short time is possible corresponding to electronic information processing. Or an effective electronic data output device or bibliographic information, such as a guide or electronic index when you want to read an electronic book between busy times, click on the required place and the part can be displayed and read after the page A survey method is obtained. In this way, the present invention can process even one piece of document information. Furthermore, since the present invention uses a search network structure, such processing of each document can be continuously processed or continuously processed. You can also

As described above, the present invention thoroughly develops a document having a desired content from a time series database obtained by combining a plurality of search modules, and develops the development ability of a certain company that is input every specific period. Many analyzes are possible, such as investigating and analyzing the performance of a specific individual based on the process of document creation. In addition, an efficient search flow can be easily established by comparing the contents of each output with the flow. In the development department, the technical rotation time for the advanced technology can be obtained in a time series, and the numerical value evaluated from the number of times cited as the prior art document can be displayed on the document map as one method of evaluating the technical ability. This particular person-created document time series database 109 or the like may be, for example, the contents of the document survey result summary table of FIGS. 5 and 9, and in particular, if the contents are patents, the claims, description summaries, etc. Parts and drawings are described in the content column, and it is easy to use if output is possible. It is desirable to include summaries and summary columns, as well as photographs and charts that are characteristic data, even in literature including humanities and economic indicators, or medical academic reports. In other words, it is desirable to access such a database containing a lot of information. In other words, it is possible to easily search a plurality of databases containing a lot of literature information as described above, and the most efficient. Thus, the present invention provides a search network structure and a search system that can obtain a proper information flow, that is, which node should be passed. The document maps and personal name lists in FIGS. 10, 11, 12, 18, and 19 are examples of output, and are not limited to the contents shown in this example. In other words, not only the number of documents, the number of creators, and the creation date, but also bibliographic items of documents such as the document publisher and the organization to which the creator belongs can be compiled and used as a reference for the survey. However, it is desirable that there is an output in which the document content or the number of documents can be checked at a glance in time series. Naturally, the summary of these document information does not have to be one output.

  As described in ST42 in FIG. 7 and the like, if the system is capable of consulting with the expert of the survey, the requesting side simply sends an email with the purpose of the search and the specific items to be searched, and the desired output contents. The search flow in FIG. 13 is performed by the survey side, and a reply can be obtained by attaching an email. If the requester wants to display a document when setting similar terms or dissimilar terms, the researcher may change the search specific items and threshold settings again to re-investigate the request to narrow down or expand the content. I can do it. Alternatively, a general server may connect to the network 3 and investigate from the distributed program. This makes it easy to connect to databases that are broad and rich, such as databases related to general literature and databases established in foreign countries. In that case, the program distribution side collects admission fees for downloads or advertising costs. It is possible to create a system that can provide answers at a reasonable cost each time a consultation is made about the contents and scope of the survey, or a membership system. Of course, you can listen to the advertising costs and make it free.

  If there is such an application software provider, you can freely search for students of universities, high schools, elementary and junior high schools, researchers at research institutes, or any individual who want to know, regardless of their academic background This makes it possible to use a database effectively at a low cost, and further improve the progress of research and the effectiveness of information exchange.

  As described above, the present invention selects document information including personal information that is information of a document creator, a specific individual selecting unit that selects a specific individual who created a document included in the document information, and a specific individual selecting unit. A plurality of document information created by the specified individual and output together on a time-series axis, and output summary means for outputting the personal information of the specified individual, the specified individual is included in the document information, and the document is created Including information related to the creation of a document by a person, an information research apparatus that can be easily narrowed down and does not require time and effort is obtained.

  As described above, the present invention extracts specific terms from sentences included in the document information, and classifies similar documents in a plurality of documents based on the appearance frequency of the specific terms, or is similar in sentences. Classifying similar terms or extracting extracted terms that are specific terms in a distribution range with a high frequency of appearance, and extracting similar terms from ranges that do not include the distribution range from which specific terms are extracted Since there is a dissimilar term extracting means for extracting extracted terms and an output summary means for outputting a similar document group or similar terms or extracted terms and dissimilar terms or non-extractable terms in a displayable manner, the content of the document information is An easy-to-understand information survey device can be obtained.

  As described above, the present invention is a similar document in which a specific term is extracted from document information including personal information which is information of a document creator and a sentence included in the document information, and a similar document group is classified from the appearance frequency of the specific term. Contents of the document information obtained by the output summarizing means, comprising: a group classification means; and an output summarizing means for summarizing the specific literature information classified as similar documents by the similar document group classification means in time series Is displayed in correspondence with the time-series axis, so that an information survey device in which the contents of the document information are easy to understand can be obtained.

  As described above, the present invention extracts specific terms from sentences included in literature information, classifies similar terms that are similar in sentences based on the appearance frequency of the specific terms, or has a distribution range with a high appearance frequency. Similar classification means for extracting specific terms, non-similar term extraction means for extracting non-similar terms or non-extracted terms from a range not including the distribution range obtained by extracting specific terms extracted from sentences included in literature information, and literature A sentence classification means for classifying the sentence or figure of the document based on the content of the document information, and classifying the similar term and at least one of the extracted term, the dissimilar term, and the non-extracted term into the categories classified by the sentence classification means Since it is displayed, it is possible to make it easy to see and understand even a document article that is difficult to read and a huge amount of document information. In the present invention, the appearance frequency distribution range of the specific term extracted from the sentence included in the document information is classified as a similar document or similar term, or the appearance frequency distribution range of the specific term that classifies the extracted term is set as the most frequently occurring frequency. Information that makes it easy to judge the content because the appearance frequency distribution range for extracting dissimilar terms or non-extracted terms is equal to or less than the intermediate range and is a distribution range of multiple occurrences, with the intermediate range between zero occurrences An investigation device is obtained.

  As described above, the present invention is based on document information including personal information that is information of a document creator, specific individual selecting means for selecting a specific individual who has created a document included in the document information, and sentences included in the document information. Similar document group classification means for extracting specific terms and classifying similar document groups from the appearance frequency of the specific terms, and specific document information and document group classification means created by a specific individual selected by the specific individual selection means And the output summarizing means for collecting the specific document information classified as similar documents on a time-series axis, so that an information investigation device with easy-to-understand output contents can be obtained.

  As described above, according to the present invention, the document information collected by the output summarizing means is based on the document contents, the display of each specific individual, the number of documents, the relationship between documents in a plurality of documents, and the text included in the document information. Since at least one of the relationships between the extracted terms is displayed in two or more dimensions on the time series axis, an easy-to-understand and easy-to-understand information survey apparatus can be obtained. In the present invention, the relationship between the documents collected in the output summary means or the relationship between the terms extracted from the sentences included in the document information is similar to the contents included in the document information obtained from the documents. Information that makes it easier to understand because the meaning of the term is similar or the extracted range is different, or the dissimilar or extracted distribution ranges are different An investigation device is obtained.

  As described above, the present invention includes document information including personal information that is information of a document creator, a plurality of first search modules having a procedure for selecting a specific individual who created a document included in the document information, Output summarizing means for summarizing specific document information created by a specific individual selected by one search module on a time-series axis, and different search conditions for the plurality of first search modules by the output summarizing means. Since the same data among the plurality of data retrieved in (1) is collected so as not to be duplicated, a practical information investigation device can be obtained with a simple configuration.

  As described above, the present invention extracts a plurality of document information including document creation date information, and a similar document that extracts a specific term from sentences included in the document information and classifies a similar document group from the appearance frequency of the specific term. Since it has a second search module for performing group classification and an output summarizing means for collecting similar document groups classified by the second search module on a time series axis, an information survey device that is easy to understand with a simple configuration Is obtained. Further, the present invention provides a first search module having a procedure for selecting document information including personal information which is information of a document creator, a specific individual who has created a document included in the document information, and a procedure different from the search module. The third search module that searches for other information other than the personal information included in the reference information, the specific reference information created by the specific individual selected by the first search module, and the third search module Since the output summarizing means for summarizing the different information on the time axis is provided, an information investigation device capable of many types of processing can be obtained. Furthermore, since the search by the second search module or the third search module of the present invention is whether the terms or the word contents included in the text of the document information are similar, an easy-to-read information research apparatus can be obtained. Still further, according to the present invention, at least one of a plurality of first search modules, second search modules, and third search modules that connect between an entrance section for obtaining literature information and output collecting means is an entrance section. And the output bundling means are connected in parallel, in series, or in a mixed state of series and parallel, so that an information investigation device that can be used flexibly is obtained. Still further, according to the present invention, the first search module, the second search module, or the third search module, each of which is connected in plural between the entrance section for obtaining the document information and the output collecting means, is set in the network state. Since the connection is made, for example, a practical information research apparatus that can use a system in which a plurality of personal computers each provided with each search module are always connected via the Internet or LAN and communicate with each other can be obtained. This gives you the search results that many people want at any time and anywhere.

  As described above, the present invention is included in the literature information including literature information including personal information that is information of the creator of the literature, a specific individual selection means for selecting a specific individual who created the literature included in the literature information, and the literature information. A similar document group classifying unit that extracts a specific term from a sentence and classifies a similar document group based on the appearance frequency of the specific term, specific document information created by the specific individual selected by the specific individual selecting unit, and An output summarizing unit that summarizes document information classified as a similar document by the similar document group classifying unit on a time-series axis, and an output unit that obtains the document information from an information server or the like and the output summarizing unit Since the specific individual selecting means and the similar document group classifying means are arranged between them and connected to the network state, a practical information research apparatus that can also use a system formed by communication is obtained.

  As described above, the document information according to the present invention is obtained from one or a plurality of information servers and is information of one or a plurality of documents, and thus can be used in a wide range of applications. Also, the personal information described in the document information of the present invention is easy to use because it is included in the output summarized by the output summarizing means. Furthermore, the present invention can be made easy to understand by using a text division axis such as the number of pages, the number of chapters, the number of items, the number of sentences, etc. for dividing the document information in place of the time series axis collected by the output bundling means.

  As described above, when the search process is performed by the circuit from the entrance unit for obtaining the document information of the information survey apparatus of the present invention to the output summarizing means, the search process is performed by a plurality of search modules. A practical information survey method can be obtained. In addition, the document information of the present invention mainly flows through the circuit from the inlet section to the output bundling means, and can be searched and processed by flowing backward or in parallel in the inlet direction. A survey method is obtained. Furthermore, according to the present invention, since a specific circuit can be selected when a search process is performed by a circuit from the entrance to the output summarizing unit, an information investigation method capable of more efficient processing can be obtained.

  As described above, the present invention extracts a specific term from a sentence included in document information by selecting a specific individual who has created this document from document information for extracting information to be investigated, and this specific term. A similar classification step for similar classification based on the frequency of occurrences, and a time-series summarizing step that summarizes information classified as similar to the information on the specific documents created by the selected specific individual in time series A practical information survey method can be obtained.

  In addition, the present invention collects as a team including a co-author or co-inventor when collecting information on specific documents in a time series summarizing step for each specific individual selected in the individual selection step or for each specific individual input in advance. Instead, joint authors or joint inventors are grouped independently, so that an information research method that is easy to analyze can be obtained.

  As described above, the present invention extracts a specific term from a sentence included in literature information, and based on the appearance frequency of the specific term, similar terms that are similar in the sentence or a specific term in a distribution range having a high appearance frequency. A classification step for classifying extracted terms, extraction steps for extracting dissimilar terms or non-extracted terms from a range not including a distribution range obtained by extracting specific terms extracted from sentences included in the literature information, and a classification step The similar terms or extracted terms obtained in this way and the dissimilar terms or non-extracted terms obtained in the extraction step are grouped in the order of document creation described in the document information or for each sentence category that classifies the document information by content. The information survey method that is practical and easy to analyze can be obtained.

  As described above, the present invention provides a database and a plurality of specific individuals that describe the results of extracting specific document information created by a plurality of specific individuals selected from the document information, including the contents of the specific document information on a time series. Since a list indicating the above is recorded, a recording medium that can be used anywhere is obtained.

  The present invention includes an input means 5 for inputting a specific item for generating a search expression, an extraction result recording means 12 for recording document information extracted from a server based on the generated search expression, and the extracted and recorded document information. A specific individual selection means 11 for selecting a specific individual who is an author or an inventor, a term extraction means, and the like are provided. Output summary means 13 for collecting and outputting in time series, it is possible to take a glance at the results of the activities of individuals or related parties. Furthermore, since a plurality of individuals are collected as this individual, a more reliable development trend can be grasped. This invention does not search for specific research content by word or sentence for the purpose, but pursues the individual to narrow down the individual research content from the whole to a specific one in time series, From the pursuit of individuals and search of the research, the target is narrowed down to the whole, and further, it is possible to pursue all as a result while limiting the scope to be examined in the narrowing process. It is possible to obtain an investigation apparatus and method in which a desired procedure can be obtained reliably and efficiently by a simple procedure.

  In the present invention, the specific items input by the input means 5 can specifically include at least the names of specific individuals that can be set in advance or the technology to be investigated, particularly the contents to be investigated, and narrow the scope of the investigation. The result is easy and easy to study. Furthermore, since the present invention can form a search network structure in which various search flows are freely connected as nodes, an efficient and extremely fast search system can be obtained.

  The present invention is a network structure of a plurality of hierarchies in which various search flows are modularized for at least a subject such as a name of a specific person who can be set in advance or a technology to be searched, particularly a search target for expanding or narrowing a content to be searched. Since information can be made to flow through the entire search network structure that is freely connected as a node, and search can be performed, a search system that is efficient and extremely fast can be obtained even in a large-scale survey area.

  Since each search flow of the present invention can search for different search objects and search ranges, a search network structure is created by freely connecting various search flows with modules as nodes to create a document time series database Therefore, it is possible to provide an easy-to-use search system that can obtain output of various contents as desired. Furthermore, in the information research apparatus having this search network structure, a lot of different information is accessed, and each downloaded document information is processed in parallel or time difference processing in a module provided in parallel, or automatically suitable interval processing with time difference, More efficient search and information collection are possible. Furthermore, by conducting continuous surveys with such information survey devices connected to the net 3 at intervals, the time series database is continuously updated, and continuous information that is not shredded is always obtained and updated. It is also possible to perform analysis based on the data.

  In the present invention, the specific individual selecting means 11 selects a plurality of specific individuals in the order of the number of documents, and extracts the specific number of persons set in advance to the output summarizing means 13, so in one stage The examination range, that is, the amount of documents can be narrowed down, and the search can be simplified. In addition, we have explained in the range of a document creator, for example a joint author, as a specific individual, but as already explained, it may be other than personal information such as a name, and a pen name, A laboratory, or a collaborator other than a collaborator will also cooperate in document creation By including related persons such as those who collected experimental data, it is possible to perform reliable search work. In addition, since the specific individual in this invention includes a joint author or a joint inventor, it is easy to grasp the progress of development and research and the technical trend, and can prevent omission. In addition, the meanings of words used by individuals and peers are almost the same, and research can be conducted in time series, so that term extraction can be performed efficiently and the contents of the literature are easy to understand and research and analysis become easy. .

  In the present invention, the selection of the specific person in the specific person selection means 11 is provided so as to be repeated a plurality of times, and the specific person previously selected by the selection of the specific person in the specific person selection means 11 is excluded next time. Even if is continued, the amount of literature information can be narrowed down and the study can be facilitated. The specific individual selecting means 11 is used for grasping a document extracted by searching for a person's name from the document information selected in the technical problem, or selecting a specific person having a large number of cases regardless of a specific range, Since it is only necessary to select a specified specific person or a person related to the extracted person, the search is performed only by specifying the person, so that the search can be performed easily and reliably.

  The specific item input by the input means of the present invention is formed so as to be inputable in a plurality of hierarchies, and after searching for the specific item of the previous hierarchy, the related items are different from the problems of the technology of the previous hierarchy. The following search is performed on the specific item, which is the subject of the above, and the literature is extracted, and the specific individuals other than the selected specific individuals are selected, so that the search population can be narrowed down and the survey is simplified.

  The result summarized by the output summarizing means 13 of the present invention includes the document contents or the contents of the document information for each specific person or the number of cases together with the time series. The present invention can be summarized regardless of the type and field of literature, and shows various output contents in an easy-to-understand state, so that a useful apparatus and method that can be analyzed even by an individual can be obtained.

  Since the present invention collects personal information from documents, it is possible to determine that the output related to a specific person collected by the output grouping means is a document in which specific items are described, and for each individual, each team, or specific It is easy to distinguish research trends and development trends because each university or research institute specified in the item is grouped by several companies, and it is easy to conduct continuous surveys.

  The present invention extracts document information based on a search formula generated by a specific item related to information to be investigated, selects a specific individual who is an author or an inventor from the extracted document information, and creates the selected specific person Since the reference information is gathered in time series and a specific individual is a plurality of teams or a plurality of teams, a search can be performed with a small number of search procedures. By extracting and combining similar document groups in this procedure, output contents that are very easy to see can be obtained.

  According to the present invention, when selecting a specific individual, the number of selected persons can be limited and a search can be performed while suppressing the amount of information surveyed. In addition, when extracting the literature information of the present invention, the search is performed in a plurality of stages, and the specific items to be searched in the next stage are expanded by using terms such as higher concepts or related actions than in the previous stage. Easy to prevent. In addition, when extracting bibliographic information, the search is performed in multiple stages, and the specific items to be searched in the next stage are narrower than the previous stage using lower concept terms or combination terms, so the scope of the survey is narrowed. -Even if the number of cases is too large, it is possible to narrow down the scope and divide the survey, so that the target survey can be performed without hindering the search.

  The present invention retrieves a group of similar documents from a plurality of documents and displays a group of documents grouped together in a time-series manner, so that it is easy to understand the contents of the documents because it is possible to know how the similar documents are being published. In addition to becoming an effect, it is easy to summarize the development. Furthermore, it is possible to obtain an output that can be easily analyzed for a single piece of document information. By replacing with the number of pages instead of time series, the number of chapters and items to categorize sentences, the number of contract terms, etc., a map showing the contents of one document in the form as shown in FIG. 12 is obtained as an output. It is done. In other words, if such a search device is used as a table of contents for complicated contracts in electronic commerce or books using electronic information, it is effective as a guide when you need to process electronic documents in a short period of time or want to read electronic books in the meantime. It is. As described above, the present invention can process even one document information, and the present invention is a combination of search modules having a network structure and can be processed in a complicated manner. Even one document at a time can be batch processed, continuously processed, or continuously processed.

  The present invention makes it possible to input a plurality of specific items for generating a search formula for information to be investigated, extracts literature information based on the search formula generated by the specific item of the upper layer first, and extracts the extracted literature information Select a plurality of specific individuals who are authors or inventors from the above, summarize the literature information created by the selected specific individuals in time series, and then based on the search formula generated by the specific items of the lower layer, and Since the document information created by a specific individual excluding the specific individual already selected is extracted, an easy-to-handle search system can be obtained. Further, the amount of document information extracted at each stage or each layer of the present invention is limited by the number of specific individuals selected. Furthermore, when setting specific items in multiple layers, the lower layer specific items must be input at the same time as the upper layer specific input, or input after document extraction by the upper layer specific items. Selectable. This allows us to conduct a user-friendly survey. In addition, the requesting side of the present invention who requests an information investigation to the investigation side sets a specific item, and enables the investigation side to communicate about the search start of the next step in a plurality of steps or a lower layer in a plurality of specific items. Therefore, it is possible to investigate the contents to be investigated reliably. In addition, since the investigation side that conducts the information investigation of the present invention can communicate with the requesting side regarding the specific item before performing the next step of the multiple steps or the lower layer search within the multiple layer specific item, In addition, it is possible to investigate as requested by the user, and processing in the shortest time becomes possible.

  The present invention is intended to simplify the analysis of the document information by grouping information such as whether the personal information in the document information and the text which is another information other than the personal information are similar in time series. . Moreover, the information gathered, for example, the contents described in the time series and the personal name list can be recorded and carried on a single recording medium so that anyone can carry a computer anywhere. You can analyze it if you want. Furthermore, if the document map is displayed on a large screen, analysis and presentation will be possible at a study meeting where many people gather.

  The present invention is a recording medium that collectively records document information created by a plurality of specific individuals selected from document information extracted based on a search formula including a problem such as technology, and the collected result is a plurality of One axis is a time series axis based on the submission date or publication date of the document or the filing date of the patent application, and it is easy to search while always grasping the progress and trends. is there. Further, the axis other than the time series axis is at least one of the number of document information, the content of the document information, and a plurality of specific personal names, and an easy-to-understand document information record is obtained. These bibliographic information records can be recorded inside the information research apparatus, recorded in an external database, or recorded on a recording medium, and can be used according to usage or by use time.

  Further, according to the present invention, search and analysis can be performed by combining a plurality of personal computers connected via the network 3 or the like. Moreover, since the combination can be a search network structure as shown in FIG. 13, a large number of information that flies over the Internet is used as each server, or many documents that are published every moment are used as one server. The combined connection of this search network that uses many servers and search modules not only enables the latest analysis, but also more complicated information such as the global scale and the cosmic scale due to more specialized information. In addition, it is possible to provide a useful search system capable of investigating a complicated analysis such as a collective feedback structure in which a plurality of reactions obtained by genetic manipulation in medicine or biology are collected. This reduces the effort of researchers to obtain information, and makes it possible to use the research time more effectively and to provide a device that is useful for creating effective ideas.

It is a schematic block diagram which shows the information investigation apparatus of Embodiment 1 of this invention. It is a flowchart explaining the operation | movement of the information investigation of Embodiment 1 of this invention. It is a schematic block diagram which shows another information investigation apparatus of Embodiment 1 of this invention. It is explanatory drawing explaining the form which requests the information investigation of Embodiment 1 of this invention. It is a summary explanatory drawing explaining the information investigation result of Embodiment 1 of this invention. It is a literature information map explanatory drawing explaining the information investigation result of Embodiment 1 of this invention. It is a flowchart explaining another operation | movement of the information investigation of Embodiment 1 of this invention. It is a flowchart explaining another operation | movement of the information investigation of Embodiment 1 of this invention. It is summary explanatory drawing explaining another summary which is the information investigation result of Embodiment 1 of this invention. It is person name explanatory drawing explaining the person name list | wrist which is the information investigation result of Embodiment 1 of this invention. It is another literature information map explanatory drawing explaining the information investigation result of Embodiment 1 of this invention. It is another literature information map explanatory drawing explaining the information investigation result of Embodiment 1 of this invention. It is network structure explanatory drawing explaining the information investigation apparatus of Embodiment 1 of this invention. It is another network structure explanatory drawing explaining the information investigation apparatus of Embodiment 1 of this invention. It is another network structure explanatory drawing explaining the information investigation apparatus of Embodiment 1 of this invention. It is a schematic block diagram which shows another information investigation apparatus of Embodiment 1 of this invention. It is a flowchart explaining the operation | movement of another information investigation of Embodiment 1 of this invention. It is a summary explanatory drawing explaining another information investigation result of Embodiment 1 of this invention. It is a summary explanatory drawing explaining another information investigation result of Embodiment 1 of this invention.

Explanation of symbols

  1, 2 information survey device, 3 network, 4 control device, 5 input means, 6 display means, 7 recording medium, 8 input / output device, 9 communication means, 10 search expression generation means, 11 specific individual selection means, 12 extraction result Recording means, 13 output summarizing means, 14 result display / recording control means, 15 recording means, 21, 22 information server, 23 database.

Claims (5)

From each document searched through communication means using one or more types of search formulas, the date and document content that can be understood as the author and inventor's name and document content described in each document being completed. Recording means for extracting and recording document information of each of the documents including
Specific person selecting means for extracting the number of documents for each specific person as the person name from the document information of each document recorded in the recording means, and specifying a plurality of specific persons as the person names in order of the number of documents When,
Retrieval including the keyword from the literature information of each document retrieved by the retrieval formula including the keyword recorded in the recording means and retrieved by the retrieval formula including the keyword input from the input means connected to the recording means A first module having the specific individual selecting means for extracting the number of documents of each document retrieved by a formula and identifying a plurality of specific individuals in the order of the large number of documents ;
A search expression that includes a person name connected to the recording means and includes a person name and does not include a keyword, or a specific person specified by the specific individual selection means included in the first module is a search expression for a person name . number of documents of each document from the document information of each document that has been retrieved by the search expression that does not include the keyword with the retrieved by search expression that does not include keywords include any of the names that are recorded in the recording means as well as A second module having the specific individual selection means for extracting a plurality of specific individuals in the order in which the number of documents is large ,
Document information that is connected to the first module and the second module and is searched for in both the first module and the second module, and is specified in each of the first module and the second module. In addition, combining the documents in the same search period from the first module and the second module that have extracted the document information including the specific individual , the document information of the combined documents without overlapping the same document information Output summarizing means for combining the documents in each document included in the list in the order in which the number of documents of each document is combined and in the order of the dates;
An information survey device characterized by comprising: 2. The specific individual specified by the specific individual selection means of the first module or the second module is specified by a plurality of searches using different search formulas depending on the input of the input means. Information survey equipment. 3. The information research apparatus according to claim 1, wherein the specific individuals are specified by the specific individual selecting means included in the first module or the second module up to a preset number of people. 4. The information investigation apparatus according to claim 1, wherein the first module and the second module are connected to a network state between the recording unit and the output summarizing unit. The said 1st module and said 2nd module which connect between the said recording means and the said output consolidating means are divided and provided in the some independent apparatus, The Claim 1 thru | or 4 characterized by the above-mentioned. Information survey device.

Images