Support system, support method and support program

Abstract

We provide support systems, methods, and programs to help those seeking technical advice find appropriate researchers. [Solution] The support system comprises a learning keyword extraction unit 111 that extracts keywords from each researcher's research achievement information, a learning unit 112 that performs machine learning to associate the extracted keywords with each researcher and stores the learning results in a research information collection results and learning results DB 130, a consultation keyword extraction unit 140 that extracts keywords from the input consultation content, a matching calculation unit 141 that matches the extracted keywords with the keywords stored in the research information collection results and learning results DB 130 and calculates the degree of matching for each researcher, and a terminal output unit of the operating terminal that is a presentation unit that presents researchers with a high degree of matching calculated by the matching calculation unit 141.

Description

[Technical Field]

[0001] The present invention relates to a support system, a support method, and a support program for assisting a person who wants to consult about technology required for research or commercialization to find an appropriate researcher. [Background technology]

[0002] Industry-academia collaboration, in which universities and private companies work together, is being promoted with the aim of researching and developing new technologies and creating new businesses. Industry-academia collaboration requires finding and matching researchers with research seeds that meet the various needs of private companies.

[0003] There have been support systems that help people who want to consult about technology to find appropriate researchers (experts). Patent Document 1 discloses a work request system that searches for people or businesses who perform work at the request of others as search subjects, and makes it possible to request work from the search subject. The work request system is characterized by having: a search subject database that stores multiple pieces of information about the search subject, along with their priorities; a request content input means that has a request statement input means that allows the request content for the search subject to be entered as a text sentence and a selection means that allows the search subject to be selected; a search condition generation means that generates search conditions and search words to be used during the search based on the selected items selected by the selection means and the items entered by the request statement input means; a search means that searches for the search subject from the search subject database, taking priority into account, based on the search conditions and search words generated by the search condition generation means; an output means that outputs the search subject found by the search means and information related to the search subject in an order that takes priority into account; a request content notification means that, when the search subject is selected by the output means, sends an e-mail to the search subject notifying the search subject that they have been selected; and a request content publication means that publishes the request content entered by the request content input means in response to access from the search subject to which the e-mail was sent. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2009-251921 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention provides a support system that assists a person who wishes to consult about technology necessary for research or commercialization to accurately find a researcher who meets the needs of the person seeking advice. [Means for solving the problem]

[0006] The support system of the present invention is a support system that supports consultants who wish to consult about technology in finding appropriate researchers, and includes: a learning keyword extraction unit that extracts keywords from the research achievement information of each researcher; a learning unit that performs machine learning to associate the keywords extracted by the learning keyword extraction unit with each researcher and accumulates the learning results in a database; an input unit that inputs the consultation content; a consultation keyword extraction unit that extracts keywords from the input consultation content; a matching calculation unit that matches the keywords extracted by the consultation keyword extraction unit with keywords accumulated in the database and calculates the degree of matching for each researcher; and a presentation unit that presents researchers with a high degree of matching calculated by the matching calculation unit. [Effects of the Invention]

[0007] According to the present invention, a database can be efficiently constructed for searching for researchers who meet the needs of people seeking technical advice, and by inputting the content of the consultation in text, it is possible to accurately search for researchers who meet the needs of the people seeking advice. [Brief explanation of the drawings]

[0008] [Figure 1]FIG. 1 shows a configuration of a support system 10 according to an embodiment. [Figure 2] 1 is a block diagram showing a configuration of a support device 100 according to an embodiment. [Figure 3] FIG. 1 is a block diagram showing a configuration of an operation terminal 300 according to an embodiment. [Figure 4] 1 is a flowchart showing the operation of the learning phase according to an embodiment. [Figure 5] FIG. 1 is a diagram showing a list of research achievement information to be collected in an embodiment. [Figure 6] FIG. 1 is a diagram showing an overview of a process for performing machine learning from research achievement information according to an embodiment. [Figure 7] 1 is a flowchart showing the operation of the search phase in an embodiment. [Figure 8] FIG. 10 is a diagram showing a specific example of a process for matching with researchers in an embodiment. [Figure 9] FIG. 1 shows an example of information stored in a research information collection result and learning result DB 130 according to an embodiment. [Figure 10] FIG. 10 is a diagram showing an example of a display of researchers suited to the consultation content in the embodiment; [Figure 11] FIG. 10 is a diagram showing an example of a search screen for adding search conditions in the embodiment; [Figure 12] FIG. 1 shows the configuration of a support system 11 according to another embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described, but the present invention is not limited to the following embodiments.

[0010] (Embodiment) [1. Configuration] [1-1. Overall system configuration] The configuration of a support system 10 according to this embodiment is shown in Fig. 1. The support system 10 is made up of a support device 100, operation terminals 300 to 320, and research information DBs (databases) 200 to 220. The support system 10 utilizes information on researchers from multiple universities to search for appropriate researchers across multiple universities for clients from private companies and other organizations who wish to consult about technologies required for research or commercialization.

[0011] The support device 100 is constructed on a server and connected to be accessible from a plurality of operation terminals 300-320 via a network line such as the Internet. The support device 100 is also connected to be communicable via a network line such as the Internet to research information DBs 200-220 that store information on each researcher's research achievements such as papers, patents, and scientific research grants. The support device 100 does not need to be constructed on a single information processing device, and may be realized in a distributed form on a plurality of information processing devices.

[0012] The operation terminals 300 to 320 are installed at universities to which the researchers who can be searched belong, or at private companies that have signed a service contract for the support system 10. Users of the support system 10 use the operation terminals 300 to 320 to perform various operations. Although three operation terminals are shown in FIG. 1, the number of operation terminals is not limited to three and may be any number.

[0013] The research information DBs 200 to 220 are the Grants-in-Aid for Scientific Research database, the Researchmap (registered trademark) database, and a researcher directory. The above databases are only examples, and other databases may be used as long as they store research achievement information such as researchers' papers, and the number of databases is not limited to three and may be any number. The research information DBs may be stored within the support device 100.

[0014] [1-2. Configuration of the support device] Figure 2 is a block diagram showing the configuration of the support device 100. The support device 100 has a learning phase functional unit that learns keywords that characterize each researcher based on research achievement information such as papers for each researcher, and a search phase functional unit that extracts keywords from the content of the consultation and selects appropriate researchers based on the extracted keywords.

[0015] The support device 100 has, as functional parts of the learning phase, a communication part 120 that communicates with the research information DBs 200-220 and the operation terminals 300-320 etc. via a network line, a research information collection part 110 that uses the communication part 120 to collect research achievement information such as papers for each researcher from the research information DBs 200-220, a learning keyword extraction part 111 that extracts keywords from the collected papers etc., a learning part 112 that performs machine learning to associate the extracted keywords with each researcher, and a research information collection results and learning results DB (database) 130 that stores the results obtained by the research information collection part 110 and the results learned by the learning part 112.

[0016] The communication unit 120 communicates with other information processing devices connected to a network line such as the Internet via a network line. The communication may be in the form of either a wireless line or a wired line, and any communication method may be used.

[0017] The research information collection unit 110 collects research achievement information such as papers for each researcher via the communication unit 120. The research achievement information is a history of the research activities that the researcher has conducted up to that point, including published papers, presentations at research meetings, etc., patents applied for as an inventor, information on applications for scientific research grants, and the details of collaborative research. It also includes past performance in responding to technical consultations. The research information collection unit 110 acquires the researchers for whom research achievement information is to be collected from the research information collection results and learning result DB 130, and collects research achievement information related to those researchers. The results acquired by the research information collection unit 110 are registered in the research information collection results and learning result DB 130.

[0018] The learning keyword extraction unit 111 extracts keywords that characterize the acquired paper, etc. from the paper, etc. The learning keyword extraction unit 111 uses morphological analysis. Morphological analysis is a string extraction method that divides a paper, etc. written in a natural language, into morphemes, which are the smallest units of meaningful expression elements, and classifies them as parts of speech. The learning keyword extraction unit 111 supports not only Japanese but also English, and extracts Japanese or English keywords and registers them in a keyword DB (database) 131. The learning keyword extraction unit 111 refers to the keyword DB 131 to determine words to be excluded as keywords and removes the words to be excluded from the extracted keywords.

[0019] If a keyword contains numbers, using those numbers as is can make matching difficult in the search phase. Therefore, the learning keyword extraction unit 111 appropriately replaces the numbers contained in the keyword. For example, if a paper or other document mentions a temperature condition of 15°C or higher and the keyword is learned using "15°C or higher," and the consultation content mentions a temperature condition of 20°C or higher and the keyword is changed to "20°C or higher," matching between the two will not be possible because the numbers are different. To avoid this situation, if a keyword contains numbers, the numbers are replaced by zeros, for example.

[0020] The keyword DB 131 has a dictionary that allows Japanese keywords and English keywords to be associated with each other. The keyword DB 131 also stores, among the keywords extracted from papers, etc., words that are inappropriate as keywords that characterize the papers, etc. (keyword exclusion words). Keyword exclusion words include, for example, the particles "wa" and "ni," the conjunctions "mata" and "shikashi," and nouns that are inappropriate as keywords. Keyword exclusion words are not limited to Japanese words but also include English words. Keyword exclusion words in the keyword DB 131 are added or deleted as appropriate by operations on the operating terminal.

[0021] The learning unit 112 uses a Bayesian filter. A Bayesian filter is a machine learning algorithm used to identify spam emails, etc., and classifies natural language documents expressed with a huge number of word combinations. The Bayesian filter identifies words that are decisive for classification and scores that indicate the strength of those words as decisive factors, and records them in a classification dictionary. The learning unit 112 identifies keywords extracted from research achievement information such as papers and the frequency of their appearance for each researcher, learns appropriate keywords that characterize each researcher, and stores them as a classification dictionary in the research information collection results and learning result DB 130.

[0022] The research information collection results and learning result DB 130 stores the learning results learned by the learning unit 112 as well as researcher information for researchers at each university who can handle the content of the technical consultation. The researcher information is information for identifying an individual researcher, such as the researcher's name, the university and faculty / department to which they belong, and the researcher's ID. The researcher information stored in the research information collection results and learning result DB 130 is registered and updated by a person with specific authority (hereinafter referred to as the administrator), such as a coordinator who promotes industry-academia collaboration at each university, using an operation terminal installed at the university. Researcher information can only be registered and updated by the administrator of the university to which the researcher belongs.

[0023] The support device 100 has, as functional parts of the search phase, a consultation keyword extraction unit 140 that extracts keywords from the content of the input technical consultation, and a matching calculation unit 141 that calculates the degree of matching between the extracted keywords and each researcher.

[0024] The consultation keyword extraction unit 140 extracts keywords that characterize the technical consultation content from the content of the consultation. The consultation keyword extraction unit 140 uses morphological analysis to extract keywords. Like the learning keyword extraction unit 111, the consultation keyword extraction unit 140 supports not only Japanese but also English, and refers to the keyword DB 131 to extract keywords excluding keyword exclusion words. When the extracted keyword is Japanese, the consultation keyword extraction unit 140 identifies the corresponding English keyword and outputs it as a pair with the Japanese keyword. When a keyword contains numbers, the consultation keyword extraction unit 140 replaces the numbers by, for example, changing all of the numbers to zero.

[0025] The matching calculation unit 141 uses a Bayesian filter. The Bayesian filter divides the document to be classified into words and compares each divided word with a classification dictionary. When the Bayesian filter finds a word in the classification dictionary, it adds it to the judgment score, calculates judgment scores for all words contained in the document to be classified, and classifies the target document based on the judgment score values. The matching calculation unit 141 compares the keywords extracted by the consultation keyword extraction unit 140 with the research information collection results and learning result DB 130 and calculates the degree of matching with each researcher.

[0026] The support device 100 is composed of a memory and a processor, and each functional unit constituting the support device 100 is realized by software that operates in cooperation with the memory and the processor. The support device 100 may also be realized as a program that runs on a computer. Note that the method for realizing each functional unit is not limited to software, and some may also be realized by hardware.

[0027] [1-3. Configuration of the operation terminal] 3 is a block diagram showing the configuration of the operation terminal 300. The operation terminal 300 includes a terminal input unit 301 for performing various input operations such as inputting consultation details, a terminal output unit 302 for outputting the screen displayed during operation and the operation results, a terminal control unit 303 for controlling the entire operation terminal 300, and a terminal communication unit 304 for communicating with the support device 100 via a network line. The operation terminal 300 is an information processing device such as a personal computer that is configured with a memory and a processor. The operation terminals including the operation terminals 310 and 320 have the same configuration as the operation terminal 300.

[0028] The terminal input unit 301 is made up of a mouse and keyboard for performing input operations on the operation terminal 300, and software for processing the received input operations. The terminal output unit 302 is composed of an output device such as a display or a printer, and software for controlling the output device. The terminal control unit 303 is composed of a memory, a processor, and software. The terminal communication unit 304 communicates with other information processing devices connected to a network line such as the Internet via a network line. The communication may be either a wireless line or a wired line, and any communication method may be used.

[0029] [2. Operation] The operation of the support system 10 configured as described above will be explained below. The operation of the support system 10 consists of a learning phase in which keywords that characterize each researcher are learned based on research achievement information such as papers for each researcher, and a search phase in which keywords are extracted from the consultation content and appropriate researchers are selected based on the extracted keywords. First, the operation of the learning phase of the support system 10 will be explained.

[0030] [2-1. Learning Phase Operation] A flowchart of the operation of the learning phase of the support system 10 is shown in Figure 4. The processing of the learning phase can be executed only by the overall administrator who manages the entire support system 10 and the administrators of each university. As an example, a case where the overall administrator executes the processing of the learning phase using the operation terminal 300 will be described.

[0031] The administrator of each university registers the researcher information of researchers belonging to their university in advance in the support device 100 using an operation terminal installed at the university. The registered researcher information is accumulated in the research information collection results and learning result DB 130. The administrator periodically updates the researcher information of researchers belonging to their university.

[0032] Some of the research information DBs 200-220 also contain researcher information, but if this information is used, incorrect researcher information will be used because the research information DBs 200-220 will not be updated if the researcher moves to another university.By having the administrator of each university periodically update the information on researchers enrolled at their university, it becomes possible to correctly determine which university each researcher belongs to.

[0033] The overall administrator uses the operation terminal 300 to instruct the support device 100 to collect research achievement information. The research information collection unit 110 identifies the target researcher from the researcher information stored in the research information collection results and learning result DB 130 (S400). The research information collection unit 110 accesses the research information DBs 200-220 via the communication unit 120 and collects research achievement information such as papers of the target researcher (S401). An example of the research information DB accessed by the research information collection unit 110 and the research achievement information collected from that database is shown in FIG. 5. Note that the research information DB accessed by the research information collection unit 110 and the research achievement information collected are not limited to those shown in FIG. 5; any information may be collected from any database as long as it represents the achievements of the research activities that the researcher has conducted up to that point.

[0034] The learning keyword extraction unit 111 extracts keywords contained in the collected papers, etc. (S402). The learning keyword extraction unit 111 extracts keywords using morphological analysis. The learning keyword extraction unit 111 outputs Japanese and English keywords. The learning keyword extraction unit 111 identifies keyword exclusion words by referring to the keyword DB 131, and outputs keywords obtained by excluding the keyword exclusion words from the extracted keywords. If a keyword contains numbers, the learning keyword extraction unit 111 replaces the numbers by, for example, changing all of the numbers to zero.

[0035] For the target researcher, research achievement information such as related papers is collected sequentially and keywords are extracted (if yes in S403). Once all research achievement information related to the researcher has been collected from the database being searched and keywords have been extracted (if no in S403), the next process will be performed.

[0036] The learning unit 112 uses a Bayesian filter to identify keywords extracted from research achievement information such as papers for the target researcher and the frequency of appearance of the keywords, calculates a score for each keyword according to the frequency of appearance (S404), and stores the results as a classification dictionary in the research information collection results and learning result DB 130 (S405). The learning unit 112 uses the many keywords extracted from papers etc. to learn which keywords should be used as keywords that characterize each researcher, and with what weighting, and stores the learning results in the research information collection results and learning result DB 130.

[0037] Specifically, an example of the research achievement information collected for Researcher A and the information recorded in the research information collection results and learning result DB 130 is shown in Figure 6. The research information collection unit 110 collects, as research achievement information, an outline of the grant-in-aid for scientific research from the Grant-in-Aid for Scientific Research database, papers from the Researchmap (registered trademark) database, and paper abstracts from the researcher directory. The research information collection results and learning result DB 130 records keywords extracted from the collected research achievement information, the number of times they appear, and a score for each keyword calculated based on the number of times they appear.

[0038] In this way, by using a Bayesian filter to extract and weight keywords from each researcher's research achievement information, such as papers, and creating a classification dictionary for each researcher, it is possible to efficiently build a database for searching for researchers who meet the needs of the client.

[0039] Furthermore, when building the research information collection results and learning results DB 130, in addition to publicly accessible databases, the content of technical consultations received in the past from clients of private companies and other organizations, which are managed by each university, and information on the researchers who actually responded to those consultations are used as learning data. This makes it possible to build a more accurate research information collection results and learning results DB 130.

[0040] The operations of S400 to S405 are performed sequentially for each researcher whose researcher information is in the research information collection results and learning results DB130 (if yes in S406), and when processing has been completed for all researchers (if no in S406), the learning phase ends.

[0041] [2-2. Search Phase Operation] The operation of the support system 10 in the search phase will be explained using the flowchart shown in FIG. 7. The processing in the search phase can be performed not only by the overall administrator and administrators of each university, but also by consultants belonging to private companies who have signed a service contract with the support system 10 and wish to consult about technologies required for research or commercialization. As an example, we will explain a case where a consultant inputs the details of a consultation using an operation terminal 300 installed at University A and an appropriate researcher is presented. FIG. 7 shows the operations of the operation terminal 300 and the support device 100 that constitute the support system 10.

[0042] A user of the support system 10 inputs the consultation content using the terminal input unit 301 of the operation terminal 300 (S700). The terminal control unit 303 of the operation terminal 300 transmits the input consultation content to the support device 100 using the terminal communication unit 304. The consultation content document can be input in English as well as in Japanese.

[0043] The consultation keyword extraction unit 140 of the support device 100 extracts keywords included in the consultation content received via the communication unit 120 (S701). The consultation keyword extraction unit 140 extracts keywords using morphological analysis. If the extracted keywords are in Japanese, the consultation keyword extraction unit 140 identifies corresponding English keywords and outputs them as pairs with the Japanese keywords. If the input consultation content is written in English, the consultation keyword extraction unit 140 extracts English keywords. The consultation keyword extraction unit 140 identifies keyword exclusion words by referring to the keyword DB 131 and outputs keywords obtained by excluding the keyword exclusion words from the extracted keywords. Note that in this embodiment, the consultation keyword extraction unit 140 identifies excluded words by referring to the keyword DB 131, similar to the learning keyword extraction unit 111. However, different keyword exclusion words may be used during learning and during searching. In this case, the keyword DB 131 separately manages keyword exclusion words for learning and keyword exclusion words for searching. If the keyword contains numbers, the consultation keyword extraction unit 140 replaces the numbers by, for example, changing all of the numbers to zero.

[0044] The matching calculation unit 141 uses a Bayesian filter to compare the keywords extracted by the consultation keyword extraction unit 140 with the research information collection results and learning result DB 130 and calculates the degree of matching with each researcher (S702). If the input consultation content is in English, the matching calculation unit 141 calculates the degree of matching between the extracted English keywords and the English keywords of each researcher stored in the research information collection results and learning result DB 130.

[0045] A specific example of calculating the degree of matching will be described with reference to FIGS. 8 and 9. FIG. 9 shows an example of data stored in the research information collection result and learning result DB 130, in which the learning results of Researcher A and Researcher B are recorded. FIG. 8 shows a specific example of calculating the degree of matching for the content of an input technical consultation. In FIG. 8, the consultation keyword extraction unit 140 extracts keywords ("fever," "cycad," "thrips," ...) from the input technical consultation content ("Understanding the mystery of fever-generating cycads. Cycads are..."). As shown in FIG. 8, the matching calculation unit 141 compares each extracted keyword with the learning results of Researcher A and Researcher B and calculates a score representing the degree of matching between Researcher A and Researcher B. In the example of FIG. 8, Researcher A's score is higher than Researcher B's score. Therefore, it can be determined that Researcher A is more suitable than Researcher B as a researcher to handle the content of the input technical consultation. The matching calculation unit 141 calculates the scores between the keywords extracted from the consultation content and all researchers stored in the research information collection results and learning result DB 130, and extracts researchers with high scores. The matching calculation unit 141 transmits the information of the extracted researchers to the operation terminal 300 using the communication unit 120.

[0046] In this way, by calculating a score based on the degree of match between the keywords extracted from the consultation content and the researchers stored in the research information collection results and learning results DB130, and extracting researchers with high scores, it is possible to accurately search for appropriate researchers who meet the needs of the person seeking advice.

[0047] The operation terminal 300 presents the received researcher information to the user of the support system 10 using the terminal output unit 302 (S703). When presenting researchers, the terminal output unit 302 displays them in order of the degree of matching. An example of a display screen by the terminal output unit 302 is shown in FIG. 10. In FIG. 10, the display screen displays recommended researchers in the order of highest score indicating the degree of matching: Researcher A, Researcher B, and Researcher C. By checking the display screen, the user of the support system 10 can identify researchers suitable for the content of the technical consultation they have entered. In addition to displaying the content on the display screen, the terminal output unit 302 can also output the entered consultation content and the researcher information selected from the presented information as electronic data. Outputting the data as electronic data allows it to be viewed on a general-purpose information processing device or used on another operation terminal.

[0048] Users of the support system 10 can add search conditions to the displayed list of researchers to perform a more detailed search. Fig. 11 shows an example of a screen for adding search conditions on the operation terminal 300. When search conditions are entered and "Execute" is pressed on the search screen, the operation terminal 300 uses the terminal communication unit 304 to send the search conditions to the support device 100. The support device 100 executes processing that matches the received search conditions and sends the results to the operation terminal 300. The operation terminal 300 displays the received results using the terminal output unit 302.

[0049] When a keyword is added to the "Contains Keyword" input field on the search screen, the support device 100 limits the search to researchers who have sentences that include the input keyword. When a keyword is added to the "Does Not Contain Keyword" input field, the support device 100 limits the search to researchers who do not have sentences that include the input keyword. For example, if a researcher is displayed at a high priority because a specific keyword is included, and that specific keyword is not important to the content of the technical consultation, more appropriate researchers can be extracted by limiting the search to researchers who do not have sentences that include that specific keyword.

[0050] When entering keywords in the "Include Keyword" or "Exclude Keyword" input fields, the display screen may be configured to display keywords extracted from the consultation content in advance so that they can be selected. This allows the user to easily enter keywords on the search screen and avoids entering keywords unrelated to the consultation content.

[0051] On the search screen, the input field for "Search Target" allows the user to select the type of research achievement information (e.g., "Paper," "Patent," "Scientific Research Grant Information," etc.). When the type of research achievement information (e.g., "Paper") is selected, the support device 100 extracts researchers from the results for each researcher stored in the research information collection results and learning result DB 130 using results based only on papers. Specifically, researchers are extracted by recalculating the display order based on the narrowing of the "Search Target." Alternatively, researchers may be extracted by recalculating the matching degree. When a year is specified in the input field for the target year, the support device 100 narrows down the results for each researcher stored in the research information collection results and learning result DB 130 to results for papers, etc. published in the specified year. When a university is specified in the input field for the "Target University," the support device 100 extracts appropriate researchers from the results for each researcher stored in the research information collection results and learning result DB 130 using results for only researchers belonging to the specified university.

[0052] [3. Effects, etc.] The support system 10 in this embodiment is a support system 10 that supports a consultant who wants to consult about technology in finding an appropriate researcher, and is equipped with: a learning keyword extraction unit 111 that extracts keywords from the research achievement information of each researcher; a learning unit 112 that performs machine learning to associate the keywords extracted by the learning keyword extraction unit 111 with each researcher and accumulates the learning results in the research information collection result and learning result DB 130; a terminal input unit 301 that inputs the consultation content; a consultation keyword extraction unit 140 that extracts keywords from the input consultation content; a matching calculation unit 141 that matches the keywords extracted by the consultation keyword extraction unit 140 with the keywords accumulated in the research information collection result and learning result DB 130 and calculates the degree of matching for each researcher; and a terminal output unit 302 that presents researchers with a high degree of matching calculated by the matching calculation unit 141.

[0053] This makes it possible to efficiently build a database using machine learning to search for researchers who meet the needs of people seeking technical advice. Furthermore, by calculating and extracting the appropriate researchers who meet the needs of people seeking advice, it is possible to search with high accuracy.

[0054] The support system 10 may also include a terminal input unit 301 that allows the user to further narrow down the researchers displayed on the terminal output unit 302 by specifying conditions.

[0055] By specifying conditions for selecting and excluding keywords, it is possible to specify keywords that more appropriately describe the content of the consultation, and to more accurately search for researchers that suit the needs of the person seeking advice.In addition, by specifying other conditions, it is possible to search for researchers that are more suitable to the various needs of the person seeking advice.

[0056] The learning unit 112 and the matching calculation unit 141 may use a Bayesian filter or other probabilistic methods.

[0057] By using a Bayesian filter to weight keywords extracted from research achievement information such as papers for each researcher and creating a classification dictionary, it is possible to efficiently build a research information collection results and learning results DB 130 for searching for researchers that suit the needs of the person seeking advice. By extracting researchers that match the content of the consultation from the research information collection results and learning results DB 130, it is possible to accurately search for appropriate researchers that suit the needs of the person seeking advice.

[0058] The researcher information in the research information collection results and learning results DB130 includes information such as the university, faculty, department, etc. to which the researcher belongs, and the list of researchers belonging to each university may be updated periodically only by an administrator with the specified authority at that university.

[0059] By having administrators of each university periodically update the researcher information of researchers belonging to their own university, it becomes possible to correctly grasp which university each researcher belongs to.

[0060] The support method in this embodiment is a support method for assisting a consultant who wishes to consult about technology to find an appropriate researcher, and includes a learning keyword extraction step (S402) for extracting keywords from the research performance information of each researcher, learning steps (S404 and S405) for performing machine learning to associate the keywords extracted in the learning keyword extraction step (S402) with each researcher and storing the learning results in the research information collection result and learning result DB130, an input step (S700) for inputting the consultation content, a keyword extraction step (S701) for extracting keywords from the input consultation content, a matching calculation step (S702) for matching the keywords extracted in the keyword extraction step (S701) with the keywords stored in the research information collection result and learning result DB130 and calculating the degree of matching for each researcher, and a presentation step (S703) for presenting researchers with a high degree of matching calculated in the matching calculation step (S702).

[0061] This makes it possible to efficiently build a database using machine learning to search for researchers who meet the needs of people seeking technical advice. Furthermore, by calculating and extracting the appropriate researchers who meet the needs of people seeking advice, it is possible to search with high accuracy.

[0062] (Other embodiments) Although the embodiments have been described above as examples of the present invention, the present invention is not limited to these and can be applied to embodiments in which modifications, substitutions, additions, omissions, etc. are made.

[0063] Therefore, other embodiments will be exemplified below. In the embodiment, a configuration has been described in which the administrator of each university accesses the research information DBs 200 to 220 and collects research achievement information that is made public in the research information DBs 200 to 220. As another embodiment, a configuration in which the authority to access a specific research information DB is limited to some administrators will be described with reference to FIG.

[0064] In FIG. 12 , the support system 11 is connected to the research information DBs 200-220 as well as the private research information DB 230. The private research information DB 230 is managed by University B, and only the administrator of University B can access the private research information DB 230. The private research information DB 230 stores research achievement information, etc., that researchers at University B have independently worked on within University B and that is not generally publicly available. When collecting research achievement information in the research information collection unit 110 of the support device 100 using the operation terminal 310, the administrator of University B collects the research achievement information from the private research information DB 230 in addition to the research information DBs 200-220. The administrator of University B adds the research achievement information from the private research information DB 230 to the research achievement information DB 230, and the results of learning in the learning unit 112 are stored in the research information collection result and learning result DB 130. The learning results stored in the research information collection result and learning result DB 130 can be used only by University B. Although non-public research achievement information cannot be utilized by other universities, each university can build its own unique DB130 of research information collection results and learning results.

[0065] In this way, only those with specific authorizations are given access to non-public information, while the learning results, such as keywords extracted from the university's non-public information, can be used by those affiliated with the university to match researchers in the same way as public information. This allows non-public information about researchers belonging to one's university to be effectively used as search information by those affiliated with the university, while maintaining its confidentiality from other universities, enabling more accurate searches for suitable researchers.

[0066] It is also possible to make available to all users the results of learning conducted by the administrator of University B using research achievement information from the non-public research information DB 230. This allows the non-public information of researchers belonging to each university to be effectively used as search information while maintaining its non-public status to anyone other than University B, enabling more accurate searches for appropriate researchers suited to the needs of the person seeking advice.

[0067] In the embodiment, a configuration has been described in which a Bayesian filter is used in the learning unit 112 and the matching calculation unit 141. The learning unit 112 performs machine learning to link research achievement information, which is learning data, to researchers, which are correct labels, and the matching calculation unit 141 may use a supervised machine learning method other than a Bayesian filter or another probabilistic method, as long as it is a method for classifying the consultation content into researchers, which are correct labels, using the machine learning results. [Explanation of symbols]

[0068] 10, 11 Support System 100 Support equipment 110 Research Information Collection Department 111 Learning Keyword Extraction Unit 112 Learning Department 120 Communications Department 130 Research information collection results and learning results DB 131 Keyword DB 140 Consultation keyword extraction unit 141 Matching calculation unit 200~220 Research information DB 230 Private research information DB 300~320 Operation terminal 301 Terminal input section 302 Terminal Output Unit 303 Terminal control unit 304 Terminal communication unit

Claims

1. A support system that helps people who want to consult about technology to find appropriate researchers. a learning keyword extraction unit that extracts keywords from the research achievement information of each researcher; a learning unit that performs machine learning to associate the keywords extracted by the learning keyword extraction unit with each researcher and accumulates the learning results in a database; an input section for inputting the content of the consultation; a keyword extraction unit that extracts keywords from the input consultation content; a matching calculation unit that matches the keywords extracted by the keyword extraction unit with keywords stored in the database and calculates the degree of matching for each researcher; a presentation unit that presents researchers with high matching degrees calculated by the matching calculation unit; A support system with

2. 2. The support system according to claim 1, further comprising a screen operation unit that enables a user to further narrow down the researchers presented by said presentation unit by specifying conditions.

3. The assistance system according to claim 1 , wherein the learning unit and the matching calculation unit use a Bayesian filter.

4. The research achievement information in the database includes both public and private information, and only those with the appropriate authority who belong to the group that provided the private information can access the private information, extract keywords from the private information, and perform machine learning to associate the extracted keywords with each researcher.

2. The support system of claim 1, wherein the matching calculation unit matches keywords extracted from the consultation content with keywords extracted from public information and private information stored in the database, regardless of whether the researcher has the specified authority, and calculates the degree of matching for each researcher.

5. The support system of claim 1, wherein the researcher information in the database includes information on the groups to which the researchers belong, and the list of researchers belonging to each group is periodically updated only by a person who has the specified authority and belongs to the group.

6. A support method for helping a person seeking technical advice to find an appropriate researcher, a learning keyword extraction step of extracting keywords from the research achievement information of each researcher; a learning step of performing machine learning to associate the keywords extracted in the learning keyword extraction step with each researcher and storing the learning results in a database; An input step to input the consultation content, a keyword extraction step of extracting keywords from the input consultation content; a matching calculation step of matching the keywords extracted in the keyword extraction step with keywords stored in the database and calculating the degree of matching for each researcher; a presentation step of presenting researchers with high matching degrees calculated in the matching calculation step; A support method that includes:

7. A support program that causes a computer to execute a support method for supporting a person who wants to consult about technology to find an appropriate researcher, a learning keyword extraction step of extracting keywords from the research achievement information of each researcher; a learning step of performing machine learning to associate the keywords extracted in the learning keyword extraction step with each researcher and storing the learning results in a database; An input step to input the consultation content, a keyword extraction step of extracting keywords from the input consultation content; a matching calculation step of matching the keywords extracted in the keyword extraction step with keywords stored in the database and calculating the degree of matching for each researcher; a presentation step of presenting researchers with high matching degrees calculated in the matching calculation step; A support program to help implement this.

Images