KR20230134711A - Researcher matching device, matching method and computer program for industry-university collaboration project - Google Patents

Abstract

The present invention relates to a matching device that connects researchers and laboratories based on research fields, to researchers in research technology fields (hereinafter referred to as "technology fields") and research subject fields (hereinafter referred to as "subject fields") categorized as detailed fields. a first database storing first interest data and research history data; a second database storing second interest data and recruitment data of a laboratory for the technical field and the subject field; an extraction unit that extracts first core data from the research history data using an artificial intelligence algorithm and extracts second core data from the job offer data; For the technical field and the subject area, generate a first matrix of the researchers based on the first interest data and the first core data, and the research laboratory based on the second interest data and the second core data. a matrix generator that generates a second matrix; For each of the technical fields and the subject areas, by comparing the second matrix of the laboratory with respect to the first matrix of the researcher, or by comparing the first matrix with respect to the second matrix, the technology a classification unit that classifies the laboratory or the researcher into grades according to whether or not they match in the field or subject area; and a matching unit that compares the similarity between the first matrix and the second matrix and provides at least one of the laboratories classified in the classification unit to the researcher, or provides at least one of the researchers to the laboratory. The performance of linked projects can be further improved by matching researchers or research teams with different research areas of interest while maintaining the same core research areas and capabilities.

Description

Researcher matching device, matching method, and computer program for industry-academia collaboration projects {RESEARCHER MATCHING DEVICE, MATCHING METHOD AND COMPUTER PROGRAM FOR INDUSTRY-UNIVERSITY COLLABORATION PROJECT}

The present invention relates to a matching device, matching method, and computer program that can link researchers and labs or researchers and research teams with similar research fields.

As technology becomes specialized and advanced, researchers' research fields are branching into various specialized fields. As a result, there are cases where there are limitations in conducting research with an individual researcher or a single research team, so collaboration between researchers, research teams, or laboratories with similar research fields is required. In the case of companies that have research labs, they have a variety of research labs. However, when the laboratory wants to recruit additional researchers or research teams or collaborate, there is a problem in which appropriate matching cannot be made because the laboratory does not have the personnel authority to make actual decisions. In particular, with the recent restrictions on research due to the COVID-19 situation, it is difficult to even recruit attendants to carry out research projects. Additionally, as research fields become fragmented, it is becoming difficult to identify researchers, research teams, or labs that are conducting or have needs for a specific research field.

Accordingly, the applicant divided the research field into technology field and subject field and attempted to perform matching between researchers, research teams, or laboratories based on the similarity of the research field, and differentially applied weights to the technology field and subject field to ensure more relevance. Research and development were conducted to enable high matching.

National Open Patent WO2008/004563

The present invention seeks to provide a matching device, matching method, and computer program that can link researchers, research teams, or laboratories conducting research in similar research fields. In particular, the purpose of the present invention is to match researchers or research teams with interests, capabilities, and knowledge in different fields while matching core capabilities and interests in research fields in order to increase the performance of linked projects. do.

The problems that the present invention seeks to solve are not limited to the problems mentioned above. Other problems and advantages of the present invention will be more clearly understood by the description below.

In order to achieve the above object, the present invention provides a matching device that connects researchers and laboratories based on research fields, with research technology fields classified into detailed fields (hereinafter referred to as "technology fields") and research subject fields (hereinafter referred to as " a first database storing the researcher's first interest data and research history data for the 'subject field'); a second database storing second interest data and recruitment data of a laboratory for the technical field and the subject field; an extraction unit that extracts first core data from the research history data using an artificial intelligence algorithm and extracts second core data from the job offer data; For the technical field and the subject area, generate a first matrix of the researchers based on the first interest data and the first core data, and the research laboratory based on the second interest data and the second core data. a matrix generator that generates a second matrix; For each of the technical field and the subject area, the second matrix of the laboratory is compared to the first matrix of the researcher, or the first matrix is compared to the second matrix to determine the technical field. or a classification unit that classifies the laboratory or the researcher into grades depending on whether there is agreement in the subject field; and a matching unit that compares the similarity between the first matrix and the second matrix and provides at least one of the laboratories classified in the classification unit to the researcher, or provides at least one of the researchers to the laboratory. One feature is that it includes ;.

Preferably, the technical field includes at least one of IoT, artificial intelligence, image processing, security, big data, mobile, automatic control technology, desktop SW, virtual reality, and blockchain, and the subject field is detailed. Classification may include at least one of non-profit/public research, profit/private, work automation/intelligence, and corporate analysis/improvement.

Preferably, the first or second data of interest includes at least one of the technical fields and at least one of the subject fields, and the first or second key data includes one of the technical fields and the It can cover any one of the subject areas.

Preferably, the extraction unit includes a text preprocessing module that preprocesses the research history data or the job offer data to generate preprocessed data; And the extracted pre-processed data is vectorized to calculate the detection frequency, and if the pre-processed data includes key words preset for a sub-field of the technical field or a sub-field of the subject field, a weight is assigned to the sub-field. It includes a core data derivation module that derives the core data, and the core data may be information about one of the subfields of the technical field or one of the subfields of the subject field.

Preferably, the extraction unit may further include a verification module that verifies whether the core data derived from the core data derivation model corresponds to one of the plurality of technical fields or the data of interest including the subject field. .

Preferably, the core data derivation module can be vectorized into a TF-IDF based module.

Preferably, the matrix includes a technology matrix for the technical field and a subject matter matrix for the subject field, and the technology matrix or the subject matrix is a subfield or subject field of the technical field to which the data of interest corresponds. n is assigned to each subfield, and the core data can be created by assigning 3n to the corresponding subfield.

Preferably, when recommending at least one of the plurality of laboratories that are matching candidates for the researcher who is a matching target, for the technical field or the subject field, the maximum value of the first matrix of the researcher corresponds to each detail It is determined based on whether the field and each subfield corresponding to the maximum value of the plurality of second matrices match, and when at least one of the plurality of researchers who are matching candidates is recommended for the laboratory that is the matching target, the technical field Or, for the subject field, it is determined based on whether each subfield corresponding to the maximum value of the second matrix of the laboratory matches each subfield to which the maximum value of the plurality of first matrices corresponds, and the technical field And the laboratories or researchers matching the detailed fields of the above subject fields are classified into group A, the laboratories or researchers matching only the detailed fields of the above subject fields are classified into group B, and the labs or researchers other than the above group A and the above group B are classified into group C. can do.

Preferably, when the matching unit recommends at least one of the plurality of laboratories that are matching candidates for the researcher who is a matching target, for the group A, the first technology matrix and the first technology matrix included in the first matrix a normalization module that normalizes one topic matrix, and normalizes the second technology matrix and the second topic matrix included in the second matrix, respectively; Delete a cell with a value of 1 among the first or second technology matrix and the first or second topic matrix, rearrange the first matrix by connecting the first technology matrix and the first topic matrix, and a rearrangement module that rearranges the second matrix by connecting a second technology matrix and the second subject matrix; and a similarity calculation module that calculates a cosine similarity for the rearranged first matrix and the plurality of rearranged second matrices.

Preferably, when the matching unit recommends at least one of the plurality of laboratories that are matching candidates for the researcher who is a matching target, for the group B, the normalization module normalizes the first and second subject matrices, and , the rearrangement module deletes and rearranges cells with a value of 1 among the first and second topic matrices, and the similarity calculation module rearranges the rearranged first topic matrix and a plurality of rearranged second topics. The cosine similarity for the matrix can be calculated.

Preferably, the matching unit may first provide the laboratory with the low cosine similarity to the group A or the group B in the similarity calculation module.

Preferably, when the matching unit recommends at least one of the plurality of laboratories that are matching candidates for the researcher who is a matching target, for the group C, the first technology matrix belonging to the first matrix and the first Connect and rearrange the subject matrix, connect the second technology matrix and the second subject matrix and rearrange the second matrix, and connect the rearranged first matrix and the rearranged plurality of second matrices. It may further include a C group judgment module that calculates cosine similarity.

Preferably, the matching unit may first provide, for the group C, the laboratory with a high cosine similarity in the group C determination module.

In addition, in order to achieve the above object, the present invention provides a matching device that connects researchers with a research team including a plurality of team members based on the research field, and the research technology field classified as a detailed field (hereinafter referred to as the 'technology field'). and a third database storing the researcher's third-party interest data and researcher history data on the research subject field (hereinafter referred to as 'subject field'); a fourth database storing fourth interest data and team member history data for the technical field and the subject field for each research team member; an extraction unit that extracts third core data from the researcher history data using an artificial intelligence algorithm and extracts fourth core data for each team member from the team member history data; For the technical field and the subject area, generate a third matrix of the researcher based on the third interest data and the third core data, and the research team based on the fourth interest data and the fourth core data. a matrix generator that generates a fourth matrix; For each of the technical fields and the subject fields, the fourth matrix of the research team is compared with the third matrix of the researcher, or the third matrix is compared with the fourth matrix to determine the technical field. or a classification unit that classifies the research team or the researchers into grades according to whether there is agreement in the subject field; and a matching unit that compares the similarity between the third matrix and the fourth matrix and provides at least one of the research teams classified in the classification unit to the researcher, or provides at least one of the researchers to the research team. Another feature is that it includes ;.

Preferably, the matrix includes a technology matrix for the technical field and a topic matrix for the subject field, and the technology matrix or the topic matrix includes the fourth data of interest for each of a plurality of team members constituting the research team. After n is given to the subfield of the technical field or the subfield of the subject field to which the fourth core data corresponds, and 3n is assigned to the subfield to which the fourth core data corresponds, the technical field and the subfield of the subject field It can be created by adding the values of all team members for .

In addition, in order to achieve the above object, the present invention provides a matching method for linking researchers and laboratories based on research fields, including research technology fields (hereinafter referred to as "technology fields") and research topic fields (hereinafter referred to as "technology fields") categorized as detailed fields. A first step of receiving the researcher's first interest data and research history data regarding the 'subject field'); A second step of extracting first core data from the research history data using an artificial intelligence algorithm; a third step of receiving secondary interest data and recruitment data from a laboratory for the technical field and the subject area; A fourth step of extracting second core data from the job offer data using an artificial intelligence algorithm; A fifth step of generating a first matrix of the researchers for the technical field and the subject field based on the first interest data and the first core data; A sixth step of generating a second matrix of the laboratory for the technical field and the subject field based on the second interest data and the second core data; For each of the technical fields and the subject areas, by comparing the second matrix of the laboratory with respect to the first matrix of the researcher, or by comparing the first matrix with respect to the second matrix, the technology A seventh step of classifying the laboratory or the researcher into one of grades A to C depending on whether there is a match in the field or subject area; And a method for comparing the similarity between the first matrix and the second matrix and providing at least one of the laboratories classified in the seventh step to the researcher, or providing at least one of the researchers to the laboratory. Another feature is that it includes 8 steps.

The present invention may be a computer program stored in a computer-readable recording medium that allows a processor to perform a matching method.

According to the present invention, synergy in research can be created by connecting researchers conducting research in a specific research field with a research team or laboratory (company) seeking to recruit or collaborate with researchers in a specific research field. Matching is possible in various forms, such as linkage between researchers and research teams, researchers and laboratories, and research teams and laboratories. In particular, the present invention allows researchers or research teams with different research areas of interest to perform linked research while matching the main core research areas and capabilities, thereby creating greater synergy and improving the performance of linked projects. It can be improved further.

The field studied by a researcher or laboratory has the advantage of high accuracy because it analyzes not only the research areas of interest that the researcher or laboratory has revealed, but also analyzes previous research results to select the most core research field among multiple research areas of interest. In the case of a research team, it is possible to derive a more accurate research field by analyzing all the research results of individual researchers and deriving a common core research field.

In addition, the research fields that serve as the basis for matching are divided into technical fields and topic fields, and weights are applied differentially. The subject area includes a classification regarding the subject or purpose of the research. Therefore, since matching takes into account not only the technical aspects of the research but also aspects of the research purpose, it has the advantage that problems regarding the direction of the research are less likely to arise even after matching.

1 is a block diagram showing the structure of a matching device according to an embodiment of the present invention.
Figure 2 is a structural diagram showing the process of calculating and matching research fields of researchers and multiple laboratories according to an embodiment of the present invention.
Figure 3 is a flowchart showing the pre-processing process performed in the text pre-processing module of the extraction unit according to an embodiment of the present invention.
Figure 4 is a reference diagram for explaining the vectorization and weighting process performed in the core data derivation module of the extraction unit according to an embodiment of the present invention.
Figures 5A to 5F are reference diagrams for explaining the process of generating a technology matrix and a subject matrix for a researcher or laboratory in the matrix generator according to an embodiment of the present invention.
Figures 6A to 6D are reference diagrams for explaining the process of generating a technology matrix and a topic matrix for a research team in the matrix generator according to an embodiment of the present invention.
Figures 7a and 7b show a matrix that can determine whether the detailed fields of the technology matrix and the subject matrix match in order to classify some of the plurality of researchers for a single laboratory into candidate groups.
Figures 8A to 8C are reference diagrams for explaining the process of selecting a recommendation priority from group A, which is a candidate for recommendation, according to an embodiment of the present invention.
Figure 9 is a reference diagram for explaining the process of selecting a recommendation priority among group B, which is a candidate for recommendation according to an embodiment of the present invention.
Figure 10 is a flowchart showing a matching method according to an embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or limited by the exemplary embodiments. The same reference numerals in each drawing indicate members that perform substantially the same function.

The purpose and effect of the present invention can be naturally understood or become clearer through the following description, and the purpose and effect of the present invention are not limited to the following description. Additionally, in describing the present invention, if it is determined that a detailed description of known techniques related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

Figure 1 is a block diagram showing the structure of a matching device 1 according to an embodiment of the present invention.

The matching device 1 can link researchers and labs, researchers and research teams, or research teams and labs based on research fields. When people with interests, capabilities, or knowledge in different areas within the same research technology field and subject area come together to conduct a linked project, synergy can be created between them, and the project can be performed better. Therefore, the factor that has the greatest impact on the success of a linking project is first bringing together people from the same research technology field and subject area. However, in general, when conducting a linked project, selecting a research field and research topic is considered the most difficult.

The matching device (1) matches the researcher-laboratory, researcher-research team, or research team-laboratory based on each interest data of the researcher, research team, or laboratory and related data about the research (research history data, team member history data, recruitment data). You can match. More specifically, matching between researchers and labs can recommend to a specific researcher a lab in a field that is most similar to the research field that the researcher is conducting among several labs. Alternatively, among multiple researchers in a specific lab, the researcher carrying out the field most similar to the research field being conducted by the lab can be recommended. In other words, in a one-to-many relationship, the most suitable organization or individual can be matched to a specific individual or organization. Here, a laboratory refers to a group of multiple people performing common research, such as a laboratory belonging to a research organization or a laboratory belonging to a company, and its form may vary. However, in one embodiment of the present invention, the description is made assuming that the laboratory is a company. Therefore, in most cases, from the standpoint of a corporate research lab, recruiting researchers or recruiting researchers for collaboration, the data that serves as the basis for matching may be ‘recruitment data.’ Here, related data about the research (research history data, team member history data, recruitment data) may be text data.

In addition, the matching device 1 can also perform matching between researchers and research teams, or between research teams and laboratories. However, here, the research team is a different concept from the laboratory. In the present invention, the laboratory has its own single direction, purpose, or field of research. On the other hand, a research team is a group of multiple individual researchers, and the direction, purpose, or field of research may not be unified. Therefore, the research team's main research field can be established by considering all relevant data regarding the research team's research and the history data of each team member. Matching between researchers and research teams can recommend a research team with the same research field to a specific researcher if the researcher wishes to belong to or collaborate with one of multiple research teams. Or, conversely, if a specific research team wants to recruit team members or collaborate, researchers with the same research field can be recommended to the specific research team. This can also be applied to matching between research teams and laboratories.

What is noteworthy is that the matching device (1) divides the research field into a more detailed research technology field (hereinafter referred to as ‘technology field’) and a research subject field (hereinafter referred to as ‘subject field’) and considers both fields. The technology field can be divided into IoT, artificial intelligence, image processing, security, big data, mobile, automatic control technology, desktop SW, virtual reality, and blockchain. Subject areas can be subdivided into non-profit/public research, profit/private, business automation/intelligence, and corporate analysis/improvement. These classifications are examples and are not limited to this, and may be classified in more detail or have a classification system. may be different. Hereinafter, ‘research field’ is defined as a general term for technical fields and subject areas.

The matching device 1 may include a database 10, an extraction unit 11, a matrix generation unit 13, a classification unit 15, and a matching unit 17.

The database 10 may store data of interest or research-related data of researchers, research teams, and laboratories in technical fields and subject areas categorized into detailed fields. Data stored in the database 10 may differ depending on the subject. That is, for researchers, data of interest and research history data can be stored, and for research teams, history data of team members who are individual researchers belonging to the research team can be stored. In the case of a laboratory, data of interest and job openings for the laboratory may be stored in the database 10.

The database 10 may include a first database 101 and a second database 103. Since the matching device 1 is a matching between a researcher and a laboratory, a researcher and a research team, or a research team and a laboratory, that is, A:B matching, the database 10 is expressed by dividing it into two groups as shown in FIG. 1. In the embodiment of Figure 1, the description is set to perform matching by recommending at least one laboratory among a plurality of laboratories to a single researcher. These settings are the same in Figure 2 below. When recommending a laboratory to a single researcher, the first database 101 may store the researcher's first interest data and research history data for the research field (technical field and subject field). The researcher's primary data of interest is information about the technical field and subject area the researcher has selected as being of interest. Data of interest may include at least one of a plurality of technical fields and at least one of a plurality of subject fields. The researcher's research history data is the results of research previously conducted by the researcher and is basic data used to calculate key data in the future. The core data calculated through analysis of the researcher's research history data is the researcher's main research field and can serve as a standard for judging similarity to the laboratory's research field.

The second database 103 may store the laboratory's second interest data and recruitment data for the research field. The laboratory's secondary data of interest is information about the field selected by the laboratory as being of interest. The second data of interest may also include at least one of a plurality of technical fields and at least one of a plurality of subject fields. However, since the researcher's data of interest and the laboratory's data of interest are different, they were separated and named data of first interest and data of second interest. Therefore, it is clarified that ‘data of interest’ hereinafter is used to collectively refer to first data of interest and second data of interest. If the subject of the data is not clearly identified as ‘data of primary interest’ or ‘data of secondary interest’, ‘data of interest’ means both data of primary interest and data of secondary interest. This also applies to core data and matrices. In other words, ‘core data’ refers collectively to the first core data and the second core data, and ‘matrix’ refers to the first matrix and the second matrix.

The laboratory's recruitment data may be data advertised by the laboratory to recruit researchers, or may include the laboratory's previous research data. The laboratory's recruitment data is basic data used to calculate key data later. The core data calculated through analysis of the laboratory's recruitment data is the main research area that reflects the main needs of the laboratory, and can serve as a standard for determining similarity to the researcher's research field. Since Figure 1 is an example of matching that recommends at least one of a plurality of laboratories to a single researcher, the second database 103 may include information on a plurality of laboratories. However, this detailed structure is arbitrary, and the database 10 may include an n-th database by forming an individual storage space for each researcher, research team, or laboratory.

The extraction unit 11 may extract first core data from research history data and extract second core data from job offer data using an artificial intelligence algorithm.

The extraction unit 11 may include a text preprocessing module 111, a core data derivation module 113, and a verification module 115.

The text preprocessing module 111 may generate preprocessed data by preprocessing research history data or job offer data.

The core data derivation module 113 can calculate the detection frequency by vectorizing the extracted preprocessed data. The core data derivation module 113 can be vectorized into a TF-IDF based module. The core data derivation module 113 may derive core data by assigning a weight to the subfield of the technical field or subject field if the preprocessed data includes key words preset for the subfield of the technical field or subject field.

Core data is information about one of the subfields of a technical field or one of the subfields of a subject field, and more specifically, may be part of data of interest. For example, if the researcher's first data of interest is IoT, image processing, automatic control technology, and desktop SW in the technology field, the first core data is one of IoT, automatic control technology, desktop SW, or image processing, and It does not apply to the excluded technology fields such as artificial intelligence, security, big data, mobile, virtual reality, and blockchain. In addition, if the researcher's primary interest data is work automation/intelligence or corporate analysis/improvement in the subject field, it is either work automation/intelligence or business analysis/improvement, and other subject areas include non-profit/public research and for-profit/improvement. It does not apply to individuals.

The verification module 115 may verify whether the core data derived from the core data derivation model 113 corresponds to any one of data of interest including a plurality of technical fields or subject fields. If, as a result of the verification, the core data does not correspond to part of the data of interest, the verification module 115 can retrain it through an artificial intelligence algorithm.

The matrix generator 13 may generate the researcher's first matrix based on the first data of interest and the first core data for the technical field and subject field, and may also generate the first matrix of the researcher based on the second data of interest and the second core data. You can create the second matrix of the laboratory. Here, the matrix can be created separately by research field. Each matrix, the first matrix and the second matrix, may include a technology matrix for the technology field and a subject matrix for the subject field, respectively. The technology matrix or topic matrix may be created by first assigning n to the subfield of the technology field or subfield of the subject field to which the data of interest corresponds, and then assigning a 3n weight to the subfield to which the core data corresponds. The matrix creation process is described in detail in Figure 5 below.

The classification unit 15 compares the second matrix of the laboratory based on the first matrix of the researcher, for each technical field and subject field, or compares the first matrix based on the second matrix to determine the technical field or subject field. Labs or researchers can be classified into levels depending on whether or not they match.

When recommending at least one of a plurality of laboratories that are matching candidates for a researcher who is a matching target, the classification unit 15 selects each subfield and a plurality of laboratories to which the maximum value of the researcher's first matrix corresponds to the technical field or subject field. It can be judged based on whether the maximum value of the second matrix matches each subfield. Afterwards, the classification unit 15 determines whether the technical field of the researcher who is the subject of matching matches the detailed field of the subject field for a plurality of laboratories that are matching candidates, and selects the laboratory that matches the detailed fields of both fields as Group A and Yang. Among the fields, labs that match only the subfields of the subject field can be classified into Group B, Group A, and labs that are not Group B can be classified into Group C. At this time, each group A, B, and C may include multiple laboratories.

Meanwhile, when recommending at least one of a plurality of researchers who are matching candidates for a matching laboratory, the classification unit 15 selects each subfield for which the maximum value of the laboratory's second matrix corresponds to the technical field or subject field. It can be determined based on whether the maximum value of the plurality of first matrices matches each subfield. Afterwards, the classification unit 15 determines whether the technical field of the laboratory and the subfield of the subject field match for the plurality of researchers who are matching candidates, and selects researchers whose subfields match in both fields into Group A and Yang. Among the fields, researchers who match only the subfields of the topic can be classified into group B, group A, and researchers who are not group B can be classified into group C. At this time, each group A, B, and C may include multiple researchers.

The matching unit 17 may compare the similarity between the first matrix and the second matrix and provide at least one of the laboratories classified in the classification unit 15 to the researcher, or provide at least one of the researchers to the laboratory. there is. The matching unit 17 may include a normalization module 171, a rearrangement module 173, and a similarity calculation module 175.

The matching unit 17 may perform matching by recommending matching candidates for Group A, Group B, and Group C classified in the classification unit 15 as matching targets in the order of Group A > Group B > Group C. What is noteworthy is that the classified candidate group differs in the preferential recommendation method within the candidate group.

Assuming that the matching unit 17 recommends at least one of a plurality of laboratories that are matching candidates for the researcher who is the matching target, for group A, the normalization module 171 selects the first technology included in the first matrix. The matrix and the first subject matrix may be normalized, and the second technology matrix and the second subject matrix included in the second matrix may be normalized, respectively.

The rearrangement module 173 deletes cells with a value of 1 among the first or second technology matrix and the first or second subject matrix, rearranges the first matrix by connecting the first technology matrix and the first subject matrix, and , the second matrix can be rearranged by connecting the second technology matrix and the second subject matrix.

The similarity calculation module 175 may calculate cosine similarity for the rearranged first matrix and the plurality of rearranged second matrices. The closer the cosine similarity is to 1, the higher the vector similarity, indicating interest in a similar field. Conversely, the closer it is to 0, the lower the vector similarity, indicating interest in different fields.

When recommending at least one of a plurality of laboratories that are matching candidates for a matching researcher, in the case of Group B, the normalization module 171 normalizes the first and second subject matrices, and the rearrangement module 173 normalizes the first and second subject matrices. And cells with a value of 1 in the second subject matrix are deleted and rearranged, and the similarity calculation module 173 can calculate the cosine similarity for the rearranged first subject matrix and the plurality of rearranged second subject matrices. there is.

In other words, the matching unit 17 can select the main matching target for group A and group B through a similar process. However, while group A considers both the technical field and the subject field, group B only considers the match between the subject fields and uses only the topic matrix. Therefore, the normalization module 171 normalizes only the topic matrix for group B, and does not use the technology matrix, so the reordering module 173 does not include a process of connecting the technology matrix and topic matrix for group B.

The matching unit 17 may further include a separate group C determination module 177 when recommending at least one of a plurality of laboratories that are matching candidates for the researcher to be matched, and in the case of group C.

The C group judgment module 177 connects and rearranges the first technology matrix and the first subject matrix belonging to the first matrix, connects the second technology matrix and the second subject matrix, rearranges the second matrix, and reorders the second matrix. Cosine similarity can be calculated for the arranged first matrix and the rearranged plurality of second matrices.

The matching unit 17 may first provide a matching candidate laboratory or researcher with a low cosine similarity calculated by the similarity calculation module 175 for group A or group B. On the other hand, the matching unit 17 may first provide a matching candidate laboratory or researcher with a high cosine similarity to group C through the group C judgment module 177.

The method of recommending matching candidates for each candidate group in the matching unit 17 will be described in detail with reference to FIGS. 6 to 8 below.

Figure 2 is a structural diagram showing the process of calculating and matching research fields of researchers and multiple laboratories according to an embodiment of the present invention. Figure 2 assumes that one of a plurality of laboratories that are matching candidates is recommended to a single researcher that is the matching target. As an example, the labs are Lab 1, Lab 2, and Lab 3. The first database 101 includes the researcher's research history data and first interest data on technology and subject areas, through which the first core data can be extracted. The second database 103 includes job opening data for each of Lab 1, Lab 2, and Lab 3, as well as secondary interest data on technology and subject areas, and is classified by assigning (1), (2), and (3). . The extraction unit 11 calculates key data for researchers and each laboratory, and can generate a matrix for technology and subject areas for each. The generated matrix can be merged into one matrix with detailed fields as columns, as shown in Figure 5e or 5f below. Afterwards, the similarity between the researcher and each laboratory can be determined through the researcher-laboratory matching algorithm, and at least one of the plurality of laboratories can be recommended.

Figure 3 is a flow chart showing the pre-processing process performed in the text pre-processing module 111 of the extraction unit 11 according to an embodiment of the present invention. The text preprocessing module 111 may generate preprocessed data by preprocessing the researcher's research history data or the laboratory's job opening data.

Steps S1101 to S1108 can be organized as in the table below.

Step S1101 is the step of converting all English letters to lowercase letters. If necessary, technical terms may be expressed in English. At this time, depending on the document writer, the first letter of a sentence or word may be expressed in capital letters, or all letters may be expressed in lowercase or uppercase letters. Therefore, unless all characters are unified into the same shape and size, there is a risk that even spellings with the same meaning will be recognized differently in a text analysis package.

Step S1102 is a synonym normalization step. Treating synonyms that indicate the same meaning but have different expression methods as the same word can reduce the occurrence of errors. In addition, there are cases where English words are written in Hangul or Korean words are written in English, and in such cases, a normalization process is performed to unify them into one to facilitate text analysis.

Step S1103 is the step where all English words are treated as nouns. In konlpy, English words are sometimes treated not as nouns but as other parts of speech. However, in Korean sites, what is usually written in English is a foreign word, that is, a noun. In addition, in order not to omit words with technically important meaning, all English words are treated as nouns and stored.

Step S1104 is a step of removing characters other than Korean and English. Special characters and numbers are often useless. For example, a word like ‘Corona 19’ can also convey meaning with just the word ‘Corona’ excluding the number. This principle ensures that all characters except English and Korean are removed. This also includes things made up of only consonants (hahaha, hahaha) that may contain typos or may be used as simple Internet buzzwords.

Step S1105 is a spell check step. Even words with the same meaning may be treated as different words by the computer if they are spelled incorrectly. For example, ‘My father is going into my bag’ and ‘My father is going into my bag’ can be treated differently. To prevent this, the Hangul notation for spacing and foreign words is unified through spell check.

Step S1106 is a morpheme classification step. For example, if you perform a morphological analysis of 'entered the bag', it is analyzed as 'bag/NNG+ eh/JKM + enter/VV+si/EPH +ㄴ다/EFN'. In this, the meaningful things are 'bag' and 'enter'. am. That is, NNG (common noun), NNP (proper noun), NNB (dependent noun), NR (rhetor), NP (pronoun), VV (verb), VA (adjective), VX (auxiliary verb), VCP (positive specifier) ), VCN (indefinite noun), etc., in part-of-speech tagging, it is nouns and verbs that have meaning in actual sentences. On the other hand, the remaining parts of speech are used to match grammatical forms or to modify sentences. Therefore, after tagging parts of speech through morphological analysis, nouns and verbs are extracted from them and those that do not have the core meaning of the sentence are excluded.

Step S1107 is a noun extraction step. Extract only nouns excluding particles, endings, etc.

Step S1108 is a stopword removal step. Even if the meaning is derived by extracting only nouns and verbs, there are words such as ‘there are’ that are used simply to match the grammatical form even though they are verbs. In this case, it is excluded because it does not have any core meaning.

Figure 4 is a reference diagram for explaining the vectorization and weighting process performed in the core data derivation module 113 of the extraction unit 11 according to an embodiment of the present invention.

The core data derivation module 113 can calculate the detection frequency by vectorizing the extracted preprocessed data.

The vectorization process is to enable text to be input to an artificial neural network after the preprocessing process performed in the text preprocessing module 111. The vectorization used at this time includes count-based vectorization and TF-IDF-based vectorization. The core data derivation module 113 can be vectorized into a TF-IDF based module. The reason is that count-based vectorization is a vectorization that only considers the frequency of simple words, so even stop words can be recognized as important, which may lead to errors. In particular, due to the characteristics of the Korean language, stop words cannot be completely removed, so TF-IDF-based vectorization is used. Vectorized preprocessed text can be input into an artificial neural network and the probability for each field can be derived by soft max regression used as an activation function.

The core data derivation module 113 may derive core data by assigning a weight to the subfield of the technical field or subject field if the preprocessed data includes key words preset for the subfield of the technical field or subject field.

Because it is difficult to completely remove stop words, TF-IDF-based vectorization was used, but this lowers the importance of frequent words for each document. However, since key words and frequently appearing words for each topic are important, text vectorization and artificial neural networks can be complemented by adding weight to the probability of each field derived in the previous step. The addition of weight is done by adding a weight of 0.07 to the probability of the field if a word corresponding to a specific word in the field exists in the input text. In the case of Figure 4, if ‘big data’, a word that frequently appears in the big data section, exists in the input document, a weight of 0.07 can be added to the big data section.

The embodiment of the present invention considers the characteristic that the performance of the artificial neural network may deteriorate as the number of hidden layers that determine the performance of the artificial neural network increases (vanishing gradient problem) and uses an RNN (Recent Layer) to compensate for this. A model can be built based on LSTM (Long Short-Term Memory model), a type of Neural Network. The model constructed in the embodiment of the present invention consists of ① an embedding layer for converting TF-IDF weights into a dense matrix (the dimension is reduced to 128 after passing through the next layer), ② an LSTM layer with 64 memory cells, and ③ a Dense layer. It can be. Soft-max can be used as an activation function.

Additionally, to adjust the weights of the artificial neural network, embodiments of the present invention can use stochastic gradient descent, and the learning-rate can be set to 0.1. Cross-entropy can be used as a loss function to reduce the difference between actual and predicted values. During training, the batch size can be set to 10 and the epoch can be set to 30. However, these detailed factors may change as the matching process progresses.

FIGS. 5A to 5F are reference diagrams for explaining the process of generating a technology matrix and a subject matrix for a researcher or laboratory in the matrix generator 13 according to an embodiment of the present invention. It is important to note that the matrices generated through the processes of FIGS. 5A to 5F are separate for each individual researcher, individual member of the research team, or individual laboratory. Figures 5A to 5F are an example of a process for generating a matrix using data of interest and core data of a specific researcher.

All data stored in the database 10 is stored in string format. However, in order to use this, the following vectorization process is required. First, you can build a zero vector (1 by the total number of detailed fields) in which each technical field and subject field are mapped. The embodiment of the present invention is a technical field divided into 10 subfields and a subject field divided into 4 subfields. Accordingly, Figure 5A is the (1 by 10) zero vector for the technical field, and Figure 5B is the (1 by 4) zero vector for the subject field.

In addition, n may be assigned to a subfield of a technical field or subfield of a subject field to which data of interest from a specific researcher, individual member of the research team, or laboratory corresponds. Figure 5c shows that n = 1 was assigned to the technical field, and it can be seen that 1 was assigned to IoT, image processing, automatic control technology, and desktop software, which are data of interest to researchers. Figure 5d shows that n = 1 was assigned to the subject subfield, and it can be seen that 1 was assigned to the data of interest to the researcher, such as work automation/intelligence and corporate analysis/improvement.

It can be determined that the core field has higher preferences and capabilities than the field of interest, and therefore, the work of adding 3n weight to the detailed field to which the core data corresponds can be carried out. The 3n weight can be set like this because 3 is the most accurate among integers. Figure 5e shows that 3n = 3 was assigned to the technical detail field, and it can be seen that 3 was assigned to the automatic control technology, which is the researcher's core data. In Figure 5f, it can be seen that 3n = 3 was assigned to the subject subfield, and 1 was assigned to corporate analysis and improvement, which is the researcher's core data.

FIGS. 6A to 6D are reference diagrams for explaining the process of generating a technology matrix and a subject matrix for a research team in the matrix generator 13 according to an embodiment of the present invention.

The matching device (1) can connect researchers with a research team including multiple team members based on the research field. It can be assumed that a researcher is the matching target and the matching candidates are multiple research teams, and the opposite case can be assumed where the research team is the matching target and the matching candidates are multiple researchers. In both cases, matching is performed in a process similar to that of FIG. 1 described above. However, in this case (researcher-research team), the names of the structure and data were differentiated for convenience to distinguish them from the researcher-laboratory matching in Figure 1.

In the case of FIG. 6A, which is an embodiment of the present invention, the third database 105 (not shown) is a configuration that stores the researcher's data, and the researcher's third interest data on technical fields and subject areas classified as detailed fields. and researcher history data can be stored. The fourth database 107 (not shown) is configured to store data of the research team, and can store fourth interest data and team member history data for technical fields and subject areas for each research team member.

What is noteworthy here is that the research team consists of multiple team members, and at this time, each team member's individual ‘team member history data’ is taken into consideration. Therefore, when extracting the fourth core data, the extraction unit 11 can extract the fourth core data for each team member using an artificial intelligence algorithm before extracting the final data of the research team.

Thereafter, the matrix generator 13 generates the researcher's third matrix based on the third interest data and the third core data for the research field, and the research team's third matrix based on the fourth interest data and the fourth core data. A fourth matrix can be created. When the research team is a matching target or matching candidate, the method of generating the matrix is somewhat different from the above-described FIGS. 5A to 5F.

For each of the plurality of team members constituting the research team, n is assigned to the subfield of the technical field or topic field to which the fourth data of interest corresponds, and 3n is assigned to the subfield to which the fourth core data corresponds. Afterwards, the fourth matrix of the research team is finally created by adding the values of all team members for the subfields of the technical field and subject field.

In more detail, Figure 6a shows that for the detailed fields of the technical field, n is assigned 1 to the fourth data of interest of each team member (team member A, team member B, and team member C), and a weight of 3n (=3) is given to the fourth core data. ) is a technology matrix given. Team member A's data of interest were given 1 each for IoT, image processing, automatic control technology, and desktop software. Since the core data extracted through the member history data of team member A in the extraction unit 11 is an automatic control technology, by multiplying the contribution of 1 by the weight of 3, the skill matrix of team member A, which is the second matrix in Figure 6a, is generated. will be. Team member B and team member C are also omitted since they are performed in the same process. Figure 6b is a topic matrix for each team member (team member A, team member B, and team member C) for the subfields of the subject area. The topic matrix can also be performed in the same manner as the process in FIG. 6A. Figures 6a and 6b show team members as rows and technical fields and subject areas as columns, respectively.

Figures 6c and 6d are the final skill matrix and topic matrix that add up all the values given to each team member for the same detailed field. In this way, in the case of a research team consisting of multiple team members, a matrix is created by considering all data of interest and core data of each team member.

For each research field, the classification unit 15 compares the fourth matrix of the research team based on the third matrix of the researcher, or compares the third matrix based on the fourth matrix to ensure agreement in the technical field or subject field. Research teams or researchers can be classified into levels depending on whether

The matching unit 17 may compare the similarity between the third matrix and the fourth matrix and provide at least one of the research teams classified in the classification unit to the researcher, or provide at least one of the researchers to the research team.

Figures 7a and 7b show a matrix that can determine whether the detailed fields of the technology matrix and the subject matrix match in order to classify some of the plurality of researchers for a single laboratory into candidate groups. If the core fields and areas of interest of the matching object (input data) and the matching candidate (recommendation object) are vectorized as shown in Figure 5 or Figure 6 described above, a process is performed to classify the plurality of matching candidates into candidate groups. Figure 7 is an embodiment of the present invention, comparing the research fields of three matching candidate researchers with respect to the laboratory that is the matching target. Figure 7a is a description matrix and Figure 7b is a topic matrix. The values of the matrices in FIGS. 7A and 7B are arbitrarily assigned values.

Candidate groups can be divided into Group A, Group B, and Group C, and the priority of recommendation is high in the following order: Candidate Group A > Group B > Group C. For example, if three or more items are not recommended in group A, additional recommendations are made in group B, and if three or more items are not recommended, the recommendations in group C can be supplemented.

The criteria for dividing each candidate group are as follows.

In candidate group A, the column with the maximum value of the matching target in the technology matrix matches the column with the maximum value of the matching candidate (and), and the column with the maximum value of the matching target (detail) in the subject matrix matches the column with the maximum value of the matching candidate (and). This is a case where the details (detailed fields) match. That is, referring to Figure 7a, the maximum value of 10 of the laboratory that is the matching target in the technology matrix is the automatic control technology. The matching candidate with the maximum value in the automatic control technology is researcher 1. Researcher 2 has the maximum value of 3 in Desktop SW, and Researcher 3 has the maximum value of 3 in IoT. And, referring to Figure 7b, the maximum value of 10 for the matching laboratory in the topic matrix is corporate analysis and improvement. The matching candidates with the highest values in the field of corporate analysis and improvement are Researcher 1 and Researcher 2. Researcher 3 has the maximum value of 3 in profit/private. Therefore, the column (subfield) with the maximum value in both the technology matrix and the subject matrix matches researcher 1. Researcher 1 belongs to group A.

Candidate group B is a case where the column (detail) with the maximum value of the matching target and the column (detail) with the maximum value of the matching candidate in the topic matrix (without considering the technology matrix) match. Since candidate B only considers the topic matrix, it is sufficient to check only Figure 7b. As described above, the maximum value of 10 for the matching laboratory in the subject matrix is corporate analysis and improvement, and the matching candidates with the maximum value in the corporate analysis and improvement field are researcher 1 and researcher 2. Since researcher 1 is included in group A, researcher 2 is included in group B.

Candidate C is a case that does not belong to Group A or Group B. In the case of Figures 7a and 7b, it is researcher 3.

Figures 8A to 8C are reference diagrams for explaining the process of selecting a recommendation priority from group A, which is a candidate for recommendation, according to an embodiment of the present invention. Figures 8a and 8b show the laboratory and researcher a or researcher to select recommendation priority among group A through similarity analysis between researcher a, researcher b, and researcher c selected as group A in the matching laboratory and the classification unit 15. It is a merge of the mattresses of c. Note that researcher a, researcher b, and researcher c are group A whose lab, column (detailed field) with the maximum value in the technology matrix and topic matrix match. Figure 8a is a description matrix and Figure 8b is a topic matrix.

This process may be performed in the normalization module 171 and the reordering module 173 in the matching unit 17. First, since the maximum values of laboratories and researchers are different, the normalization module 171 divides the values of all subfields by their respective maximum values in the technology matrix and subject matrix. Referring to Figure 8a, in the case of a laboratory, 10 technical subfields are divided into a maximum value of 10. In the case of researchers A through C, each is divided by the maximum value of 3. Since each was divided by its maximum value, the maximum values of the laboratory and researcher A through researcher C were normalized when they all became 1.

Next, in order to meet the purpose of the present invention, the reordering module 173 excludes the columns (details) to which each maximum value corresponds from the technical field and subject field. The reason is that the present invention states that linked projects can be performed better when researchers or research teams gather with people with interests, capabilities, and knowledge in different fields while the core capabilities and interests in the research field are consistent. This is because we want to make a judgment and perform such matching. Therefore, the present invention is designed to recommend subjects of interest to other fields, excluding the detailed field with the maximum value. In addition, the present invention further complements this by later recommending groups A and B in the order in which the cosine similarity is close to 0 in the similarity calculation module 175.

Then, as shown in FIG. 8C, the technology matrix of FIG. 8A and the subject matrix of FIG. 8B are combined, and the similarity calculation module 175 calculates the cosine similarity with the matching laboratory. Referring to Figure 8c, researcher a's cosine similarity with the laboratory is 0.64, researcher b's is 0.414, and researcher c's is 0.311, and the cosine similarity is in the order of researcher a > researcher b > researcher c. The closer the cosine similarity is to 1, the higher the vector similarity, indicating interest in a similar field. Conversely, the closer it is to 0, the lower the vector similarity, indicating interest in different fields. Therefore, as described above, since the present invention seeks to match groups with the same core data but different data of interest, in Figure 8c, which is group A, researcher c whose cosine similarity is closest to 0 can be recommended first.

Figure 9 is a reference diagram for explaining the process of selecting a recommendation priority among group B, which is a candidate for recommendation according to an embodiment of the present invention.

Due to the requirement that the maximum columns of the technology and topic matrices must match in Group A, it may be difficult to recommend more than 3 recommendations in Group A alone. In such cases, additional recommendations can be made from candidate B. Group B is a case where the column with the maximum value of the subject matrix of the matching target and the column with the maximum value of the matching candidate are the same. The recommendation ranking within Group B proceeds almost similarly to Group A described above in FIGS. 8A to 8C. In other words, for the purpose of the present invention, it is recommended in the order in which the cosine similarity of the reconstructed matrix is closest to 0, except for the core subject field (column with the maximum value). Therefore, in the case of Figure 9, the ‘Corporate Analysis and Improvement’ column, which is the subfield with the maximum value, is excluded from the topic matrix of group B. Then, the cosine similarity between the matching laboratory and the matching candidate researchers a to researcher c is calculated. Referring to Figure 9, the cosine similarity is in the order of researcher c > researcher a > researcher b. Therefore, in Figure 9, group B, researcher b whose cosine similarity is closest to 0 can be recommended first.

Figure 10 is a flowchart showing a matching method according to an embodiment of the present invention. The first step of the matching method that connects researchers and laboratories based on research fields can receive the researcher's first interest data and research history data on technical fields and subject areas categorized as detailed fields (S310 ). In the second step, the first core data can be extracted from the research history data using an artificial intelligence algorithm (S320). The third step may receive the laboratory's secondary interest data and recruitment data for technical fields and subject areas (S330). In the fourth step, the second core data can be extracted from the job offer data using an artificial intelligence algorithm (S340). In the fifth step, a researcher's first matrix may be created for the technical field and subject field based on the first data of interest and the first core data (S350). In the sixth step, a second matrix of the laboratory may be created for the technical field and subject area based on the second data of interest and the second core data (S360). Step 7 is to compare the laboratory's second matrix with respect to the researcher's first matrix, for each technical field and subject area, or by comparing the first matrix with respect to the second matrix. Depending on whether there is a match, a laboratory or researcher can be classified as either Grade A or Grade C (S370). The eighth step compares the similarity between the first matrix and the second matrix, and provides at least one of the laboratories classified in the seventh step (S370) to the researcher, or provides at least one of the researchers to the laboratory. (S380).

The first to eighth steps represent processes performed in the embodiment of the above-described matching device 1, and since the meaning and significance of each step has been described above, redundant description will be omitted.

Meanwhile, the matching method that connects researchers and laboratories based on the above-mentioned research fields can be implemented as an application or in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. there is. A computer-readable recording medium may include program instructions, data files, data structures, etc., singly or in combination.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specifically configured to store and perform program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include not only machine language code such as that created by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform processing according to the invention and vice versa.

Although the present invention has been described in detail through representative embodiments above, those skilled in the art will understand that various modifications can be made to the above-described embodiments without departing from the scope of the present invention. will be. Therefore, the scope of rights of the present invention should not be limited to the described embodiments, but should be determined not only by the claims described later, but also by all changes or modified forms derived from the claims and the concept of equivalents.

1: Matching device
10: database
101: first database
103: second database
105: Third database
107: 4th database
11: extraction unit
111: Text preprocessing module
113: Core data derivation module
115: verification module
13: Matrix generation unit
15: Classification section
17: Matching unit
171: Normalization module
173: Rearrangement module
175: Similarity calculation module
177: Group C judgment module

Claims (17)

In a matching device that connects researchers and laboratories based on research field,
a first database that stores the researcher's primary interest data and research history data on research technology fields (hereinafter referred to as 'technology fields') and research topic fields (hereinafter referred to as 'subject fields') classified into detailed fields;
a second database storing second interest data and recruitment data of a laboratory for the technical field and the subject field;
an extraction unit that extracts first core data from the research history data using an artificial intelligence algorithm and extracts second core data from the job offer data;
For the technical field and the subject area, generate a first matrix of the researchers based on the first interest data and the first core data, and the research laboratory based on the second interest data and the second core data. a matrix generator that generates a second matrix;
For each of the technical field and the subject area, the second matrix of the laboratory is compared to the first matrix of the researcher, or the first matrix is compared to the second matrix to determine the technical field. or a classification unit that classifies the laboratory or the researcher into grades depending on whether there is agreement in the subject field; and
a matching unit that compares similarities between the first matrix and the second matrix and provides at least one of the laboratories classified in the classification unit to the researcher, or provides at least one of the researchers to the laboratory; A matching device comprising:
According to claim 1,
The technical field includes at least one of IoT, artificial intelligence, image processing, security, big data, mobile, automatic control technology, desktop software, virtual reality, and blockchain, as a detailed classification,
The subject field is a detailed classification that includes at least one of non-profit/public research, profit/private, work automation/intelligence, and corporate analysis/improvement. A matching device.
According to claim 1,
The first or second data of interest is,
Includes at least one of the above technical fields and at least one of the above subject areas,
The first or second core data is,
A matching device comprising any one of the above technical fields and one of the above subject areas.
According to claim 1,
The extraction unit,
a text preprocessing module that generates preprocessed data by preprocessing the research history data or the job offer data; and
The extracted pre-processed data is vectorized to calculate the detection frequency, and if the pre-processed data includes key words preset for a sub-field of the technical field or a sub-field of the subject field, a weight is assigned to the sub-field. Includes a core data derivation module that derives core data,
A matching device, wherein the core data is information about one of the subfields of the technical field or one of the subfields of the subject field.
According to claim 4,
The extraction unit,
A matching device further comprising a verification module that verifies whether the core data derived from the core data derivation model corresponds to one of the plurality of technical fields or the data of interest including the subject field.
According to claim 4,
The core data derivation module is,
A matching device characterized by vectorization with a TF-IDF based module.
According to claim 1,
The matrix includes a technology matrix for the technical field and a subject matrix for the subject field,
The technology matrix or the subject matrix is:
A matching device, wherein n is given to a subfield of the technical field or a subfield of the subject field to which the data of interest corresponds, and a 3n weight is assigned to the subfield to which the key data corresponds.
According to claim 7,
When recommending at least one of the plurality of laboratories that are matching candidates for the researcher who is the matching target,
With respect to the technical field or the subject field, it is determined based on whether each subfield corresponding to the maximum value of the first matrix of the researcher matches each subfield corresponding to the maximum value of the plurality of second matrices,
When recommending at least one of the plurality of researchers who are matching candidates for the laboratory that is the matching target,
With respect to the technical field or the subject field, it is determined based on whether each subfield corresponding to the maximum value of the second matrix of the laboratory matches each subfield corresponding to the maximum value of the plurality of first matrices,
Group A refers to laboratories or researchers whose technical fields and subfields match the above subject areas;
Group B refers to laboratories or researchers that match only the detailed fields of the above subject areas;
A matching device characterized in that it classifies laboratories or researchers other than Group A and Group B into Group C.
According to claim 8,
The matching unit,
When recommending at least one of the plurality of laboratories that are matching candidates for the researcher who is the matching target, for group A,
a normalization module that normalizes the first technology matrix and the first topic matrix included in the first matrix, respectively, and normalizes the second technology matrix and the second topic matrix included in the second matrix, respectively;
Delete a cell with a value of 1 among the first or second technology matrix and the first or second topic matrix, rearrange the first matrix by connecting the first technology matrix and the first topic matrix, and a rearrangement module that rearranges the second matrix by connecting a second technology matrix and the second subject matrix; and
A matching device comprising a similarity calculation module that calculates a cosine similarity for the rearranged first matrix and the plurality of rearranged second matrices.
According to clause 9,
The matching unit,
When recommending at least one of the plurality of laboratories that are matching candidates for the researcher who is the matching target, for group B,
the normalization module normalizes the first and second topic matrices,
The rearrangement module deletes and rearranges cells with a value of 1 among the first and second subject matrices,
The similarity calculation module is a matching device characterized in that it calculates cosine similarity for the rearranged first subject matrix and the plurality of rearranged second subject matrices.
According to claim 9 or 10,
The matching unit,
A matching device, characterized in that for the group A or the group B, the similarity calculation module first provides the laboratory with the low cosine similarity.
According to claim 8,
The matching unit,
When recommending at least one of the plurality of laboratories that are matching candidates for the researcher who is the matching target, for group C,
The first technology matrix and the first subject matrix belonging to the first matrix are connected and rearranged, the second technology matrix and the second subject matrix are connected and the second matrix is rearranged, and the rearranged A matching device further comprising a C group determination module that calculates cosine similarity for the first matrix and the plurality of rearranged second matrices.
According to claim 12,
The matching unit,
A matching device, characterized in that, for the group C, the group C judgment module first provides the laboratory with a high cosine similarity.
In a matching device that links researchers with a research team including multiple team members based on research field,
A third database that stores the researcher's third-party interest data and researcher history data on research technology fields (hereinafter referred to as "technology fields") and research topic fields (hereinafter referred to as "subject fields") categorized into detailed fields;
a fourth database storing fourth interest data and team member history data for the technical field and the subject field for each research team member;
an extraction unit that extracts third core data from the researcher history data using an artificial intelligence algorithm and extracts fourth core data for each team member from the team member history data;
For the technical field and the subject area, generate a third matrix of the researcher based on the third interest data and the third core data, and the research team based on the fourth interest data and the fourth core data. a matrix generator that generates a fourth matrix;
For each of the technical fields and the subject fields, the fourth matrix of the research team is compared with the third matrix of the researcher, or the third matrix is compared with the fourth matrix to determine the technical field. or a classification unit that classifies the research team or the researchers into grades according to whether there is agreement in the subject field; and
a matching unit that compares similarities between the third matrix and the fourth matrix and provides at least one of the research teams classified in the classification unit to the researcher, or provides at least one of the researchers to the research team; A matching device comprising:
According to claim 14,
The matrix includes a technology matrix for the technical field and a subject matrix for the subject field,
The technology matrix or the subject matrix is:
For each of the plurality of team members constituting the research team, n is assigned to the subfield of the technical field or the subfield of the subject field to which the fourth data of interest corresponds, and to the subfield to which the fourth core data corresponds. After 3n is assigned, a matching device is created by adding the values of all team members for the technical field and the subfields of the subject field.
In the matching method of linking researchers and laboratories based on research field,
A first step of receiving the researcher's first interest data and research history data on the research technology field (hereinafter referred to as 'technical field') and the research subject field (hereinafter referred to as 'subject field') classified into detailed fields;
A second step of extracting first core data from the research history data using an artificial intelligence algorithm;
a third step of receiving secondary interest data and recruitment data from a laboratory for the technical field and the subject area;
A fourth step of extracting second core data from the job offer data using an artificial intelligence algorithm;
A fifth step of generating a first matrix of the researchers for the technical field and the subject field based on the first interest data and the first core data;
A sixth step of generating a second matrix of the laboratory for the technical field and the subject field based on the second interest data and the second core data;
For each of the technical fields and the subject areas, by comparing the second matrix of the laboratory with respect to the first matrix of the researcher, or by comparing the first matrix with respect to the second matrix, the technology A seventh step of classifying the laboratory or the researcher into one of grades A to C depending on whether there is a match in the field or subject area; and
8, which compares the similarity between the first matrix and the second matrix and provides at least one of the laboratories classified in the seventh step to the researcher, or provides at least one of the researchers to the laboratory A matching method comprising: a step;
A computer program stored in a computer-readable recording medium that causes a processor to perform the matching method of claim 16.