JP2021157363A - Needs matching device and program - Google Patents

Abstract

To enable efficiently finding other needs that match certain needs.SOLUTION: A needs matching device (server) comprises: a registration needs vector calculation section for, concerning a database associating and managing content sentences for a plurality of registration needs with needs classifications, based on vectors expressing each word included in the content sentences for each of the registration needs, and vectors expressing each word included in titles of the needs classifications, calculating registration needs vectors expressing the registration needs; a first reception section for receiving content sentences for inquiry needs; a search section for, based on vectors expressing each word included in the content sentences for the inquiry needs, calculating inquiry needs vectors expressing the inquiry needs, and searching for the registration needs vectors similar to the calculated inquiry needs vectors; and an output section for outputting at least the content sentences for the registration needs expressed by the searched registration needs vectors.SELECTED DRAWING: Figure 11

Description

The present invention relates to a needs matching technique.

Business in a company is made up of transactions and cooperation with other companies. As a result, companies are often looking for new business partners and partners. If a suitable partner for the company is found, business activities will proceed smoothly.

New business partners and cooperating partners may be found through personal referrals, but it is even better if there is an opportunity to meet efficiently without relying on human relationships.

It can be said that matching the needs of companies is the key to finding new business partners and cooperating partners in that transactions and cooperative relationships are established when the needs of both businesses are met. For example, if one company has a need to sell a product and another company has a need to buy that product, those needs are matched and may develop into a business relationship. Is high.

Japanese Unexamined Patent Publication No. 2018-13819

Patent Document 1 discloses a system that supports business matching. Patent Document 1 is a kind of technology for finding a partner company by creating a database of information about various companies. Information about the company is expressed in text, and the method of searching for information that matches the keyword set by the user is the basis. In other words, find matching information by comparing texts.

However, the conventional method does not always reach the desired company information. Even if they do not seem to match in terms of expression, company information that actually meets their needs may be overlooked. The amount of information about a company is enormous, and it is time-consuming and labor-intensive for a person to judge the practical suitability of needs.

The present invention is an invention completed based on the above-mentioned problem recognition, and its main purpose is to efficiently find other needs that meet one need.

The needs matching device in a certain aspect of the present invention relates to a database that manages a plurality of registered needs by associating them with a needs category, and has a vector representing each word included in each registered needs and a title of the needs category. A calculation unit that calculates a registration needs vector that represents the registration needs based on a vector that represents each word included in, a first reception unit that accepts the content sentence of the inquiry needs, and each included in the content sentence of the inquiry needs. A search unit that calculates a query needs vector that represents a query need based on a vector that represents a word and searches for a registration needs vector that is similar to the calculated query needs vector, and a registration that is represented by the searched registration needs vector. It is provided with an output unit that outputs at least the content sentence of the needs.

According to the present invention, it is possible to efficiently find other needs that meet one need.

It is a network configuration diagram of the needs matching system in this embodiment. It is a data structure diagram of the reference company table. It is a data structure diagram of a word table. It is a figure of the registration needs reception screen. It is a data structure diagram of the registration needs table. It is a data structure diagram of the needs classification table. It is a data structure diagram of the needs division vector storage part. It is a data structure diagram of the registration needs vector storage part. It is a figure of the inquiry needs reception screen. It is a figure of the matching result screen. It is a functional block diagram of a server. It is a flowchart which shows the process of the distributed representation model generation processing. It is a flowchart which shows the process of registration needs addition processing. It is a flowchart which shows the process of needs matching processing.

Business-to-business transactions and cooperative relationships are established when the needs of each company are met. If one company has a need to sell a product and another company has a need to buy that product, those needs are matched and likely to develop into a business relationship. .. If one company has a need to develop a technology and another company has a need to provide a technology related to that technology, those needs are matched and researched. And can be a development partner. Also, if one company has a need to receive a service and another company has a need to provide that service, those needs are matched and that company's service is used. May be. In this embodiment, we propose a system that supports such needs matching between companies.

FIG. 1 is a network configuration diagram of the needs matching system according to the present embodiment. In the needs matching system, needs matching is performed to match suitable needs with each other. However, instead of a text search, a search is performed by determining the similarity of vectors using a distributed representation model.

The needs matching system in this embodiment includes a server 100, a user terminal 200a, and a user terminal 200b. In this example, the user terminal 200a is used, for example, by a person in charge of a certain company to register his / her needs. Registered needs will be searched by other companies. The needs registered as search targets in this way are called "registration needs". The user terminal 200b is used, for example, when a person in charge of another company searches for a need that meets the needs of the company from among the registration needs. The needs presented in order to find suitable registration needs in this way are called "inquiry needs".

The user terminal 200a and the user terminal 200b connect to the server 100 via a network (for example, the Internet, a LAN, a dedicated line, etc.). Although the figure shows two user terminals 200, a larger number of user terminals 200 are actually used.

The server 100 has a distributed representation model. The distributed representation model is stored in the distributed representation model storage unit 130. The distributed representation model will be described below.

In the distributed representation model, many words are represented by high-dimensional real vectors. The distributed representation model is obtained by learning processing based on a database called a corpus, which is a large-scale collection of structured natural language sentences. This technique is well known in the field of natural language processing. The real number vector of a word obtained as a distributed representation model is called a "word vector". It is also called a "dispersion vector" in that it is a vector of distributed representation. It is also called a "meaning vector" in that it quantifies the meaning of text.

As a typical distributed expression technique, for example, Word2vec is known. In Word2vec, based on the premise that "words that appear in the vicinity are similar to each other are semantically close", search for another word that appears in the vicinity of the target word (for example, 5 words before and after or 10 words before and after) in the sentence. Solve the problem of doing. The five words before and after here are the range of a total of 10 words, from the word 5 before the target word to the word before 1 and the word after 1 to 5 after the target word. be. A neural network is used as the learning means. In the process of learning using a large corpus, the word vectors of words that appear in the vicinity and have similar words change to show similar values.

In the distributed representation model, not only the similarity between words can be shown by the word vector, but also the similarity between phrases, clauses, sentences and sentences can be shown. In a sentence containing a plurality of words, a dispersion vector (sometimes referred to as a "sentence vector") of the sentence or the like is generated from the word vector of these words. When the sentence vector of one sentence is similar to the sentence vector of another sentence, these sentences indicate that the contents are close to each other.

In the present embodiment, the registration needs received from the user terminal 200a are vectorized using the distributed representation model. Similarly, the matching needs received from the user terminal 200b are also vectorized. The suitability between needs is determined by the similarity of these vectors. The similarity is, for example, cosine similarity. Although the details will be described later, in the present embodiment, the vector is obtained in consideration of the needs classification regarding the registration needs.

The origin of the distributed representation model is arbitrary, but here we use an existing database of corporate information. A corpus is generated from a database of corporate information and a distributed representation model based on that corpus is used. Companies that perform needs matching do not need to be registered in this database. The company information in this database is for reference only, so it is called the "reference company database". The reference company database 120 has a reference company table in which reference company information is stored.

FIG. 2 is a data structure diagram of the reference company table.
The reference company table is only used to generate the distributed representation model and is not directly involved in needs matching. The company table has a record for each company. The record in the company table contains the reference company ID, the reference company name, and the business content. Only one sentence may be set in the field of the business content, or a plurality of sentences may be set. Alternatively, it may be a text enumerating words or phrases.

FIG. 3 is a data structure diagram of a word table.
The substance of the distributed representation model takes, for example, the data format of a word table. The word table has a record for each word. Words and word vectors are set in the records of the word table.

FIG. 4 is a diagram of the registration needs reception screen.
The registration needs reception screen is generated by the server 100 logged in from the user terminal 200a and transmitted to the user terminal 200a. Then, on the screen displayed on the user terminal 200a, information on registration needs is received from the user. In the display area 300 of the registered company, the name of the company that registers the needs is displayed. At the stage of logging in, the company name shall be specified by referring to the company table (not shown) that manages the company name, location, account information, person in charge name, e-mail address, telephone number, and so on. The content of the registration need is input in text in the input area 302 of the content of the registration need. In the needs category selection area 304, the needs category is selected. In this example, a pull-down menu method is used to select from preset needs category candidates. When the registration button 306 is touched, registration needs data (hereinafter, referred to as “registration needs data”) including the registered company name, registration needs content, and needs classification ID is transmitted to the server 100. When the cancel button 308 is touched, the registration needs reception screen is closed without transmitting the registration needs data.

FIG. 5 is a data structure diagram of the registration needs table.
The registration needs data transmitted to the server 100 is stored in the registration needs table. The registration needs table has a record for each registration need. When the server 100 receives the registration needs data, a new record is added. The registered company name, registered needs content, and needs classification ID are set in the record of the registration needs table.

FIG. 6 is a data structure diagram of the needs classification table.
The needs classification table is prepared in advance in the needs database 140. The needs classification table has a record for each needs classification. A needs classification ID and a needs classification title are set in the record of the needs classification table. The needs category title is displayed in the pull-down menu of the needs category selection area 304 described above.

FIG. 7 is a data structure diagram of the needs classification vector storage unit.
The needs classification vector storage unit is also prepared in advance in the needs database 140 according to the needs classification table. The needs classification vector storage unit has a record for each needs classification. The needs classification ID and the needs classification vector are set in association with each other in the record of the needs classification vector storage unit. The needs segment vector corresponds to the variance vector of the needs segment title. That is, the needs division vector is generated from the word vector of each word included in the needs division title. In this example, it is generated by averaging the word vector of each word included in the needs category title. For example, the needs segment vector of the title "R & D cooperation" is the average vector of the word vector of "research", the word vector of "development", and the word vector of "cooperation". Therefore, the needs classification vector of "R & D cooperation" reflects the characteristics of the word vector of "research", the characteristics of the word vector of "development", and the characteristics of the word vector of "cooperation".

FIG. 8 is a data structure diagram of the registration needs vector storage unit.
The registration needs vector is generated after the registration needs data is received by the server 100. The registration needs vector storage unit has a record corresponding to the registration needs, and the registration needs ID and the registration needs vector are set in association with each other in the record. The registration needs vector corresponds to the variance vector of the registration needs. However, it is generated by taking into account not only the registered needs content but also the needs classification. Specifically, it is generated from the variance vector of the registered needs content (hereinafter referred to as "needs content vector") and the needs classification vector. In this example, the registered needs vector is calculated by the weighted average of the needs content vector and the needs division vector. It is assumed that the weight ratio is constant. For example, if the weight of the needs content vector is 7 and the weight of the needs division vector is 3, then for each component of the vector, the value of the needs content vector is multiplied by 7 and the value of the needs division vector is multiplied by 3. The value of the registration needs vector is calculated by dividing the sum of them by 10.

For example, the registration needs vector of the registration needs ID: N0300 in FIG. 5 is obtained by the load average of 7: 3 of the needs content vector of "I want to increase the application fields of silent motors" and the needs classification vector of "R & D cooperation". .. Needs content vector of "I want to increase the application field of silent motor" is the word vector of "quiet", the word vector of "motor", the word vector of "application", the word vector of "field", the word vector of "increase". And the average vector of the word vectors of "want". Here, particles are ignored. The needs classification vector of "research and development cooperation" is the average vector of the word vector of "research", the word vector of "development", and the word vector of "cooperation". Therefore, the registration needs vector of the registration need ID: N0300 includes the feature of the word vector of "quiet", the feature of the word vector of "motor", the feature of the word vector of "application", the feature of the word vector of "field", and so on. It reflects the "increase" word vector features, the "want" word vector features, the "research" word vector features, the "development" word vector features, and the "cooperation" word vector features. A simple average may be used instead of the weighted average.

FIG. 9 is a diagram of the inquiry needs reception screen.
The inquiry needs reception screen is generated by the server 100 logged in from the user terminal 200b and transmitted to the user terminal 200b. Then, on the screen displayed on the user terminal 200b, information on inquiry needs is received from the user. The content of the inquiry need is input in text in the input area 310 of the content of the inquiry need. When the execution button 312 is touched, the inquiry needs content is sent to the server 100 as inquiry needs data (hereinafter referred to as "inquiry needs data"). When the cancel button 314 is touched, the inquiry needs reception screen is closed without transmitting the inquiry needs data.

When the server 100 receives the query needs data, it generates a query needs vector based on the text of the query needs content. That is, the query needs vector corresponds to the variance vector of the query needs. Specifically, it is generated from the word vector of the words included in the inquiry needs content. In this example, the average of those word vectors is the query needs vector.

For example, the query needs vector for "help me develop a quiet vacuum cleaner" is the word vector for "quiet", the word vector for "vacuum cleaner", the word vector for "development", the word vector for "help" and the word vector for "help". Generated from the word vector "I want". Therefore, the inquiry needs vector of "Please help develop a quiet vacuum cleaner" includes the characteristics of the word vector of "quiet", the characteristics of the word vector of "vacuum cleaner", the characteristics of the word vector of "development", and ". It reflects the characteristics of the word vector of "help" and the characteristics of the word vector of "want".

FIG. 10 is a diagram of the matching result screen.
On the matching result screen, the registration needs that match the inquiry needs are displayed in a list format. The matching result screen is returned to the user terminal 200b, which is the source of the inquiry needs data. Each registration needs line includes a registration needs content display area 320, a needs category display area 322, a registered company display area 324, and a detail button 326. When the detail button 326 is touched, detailed information about the registered company (company name, location, name of person in charge, e-mail address, telephone number, etc.) is displayed.

The registration needs displayed in the first line will be described. As explained in relation to FIG. 8, the registration needs vector generated from the needs content vector "I want to increase the application fields of silent motors" and the needs classification vector "Research and development cooperation" includes the characteristics of the word vector "quiet". , "Motor" word vector features, "Applied" word vector features, "Field" word vector features, "Increase" word vector features, "I want" word vector features, "Research" It reflects the characteristics of the word vector, the characteristics of the "development" word vector and the characteristics of the "cooperation" word vector. On the other hand, as explained in relation to FIG. 9, the inquiry needs vector of "I want you to help develop a quiet vacuum cleaner" includes the characteristics of the word vector of "quiet" and the characteristics of the word vector of "vacuum cleaner". It reflects the characteristics of the "development" word vector, the "help" word vector, and the "want" word vector.

Comparing the two, the characteristics of the "quiet" word vector of the registration needs vector are close to the characteristics of the "quiet" word vector of the query needs vector. The characteristics of the "motor" word vector of the registration needs vector are related to the characteristics of the "vacuum cleaner" word vector of the query needs vector. The characteristics of the word vector of "development" of the registration needs vector are common to those of the query needs vector. The characteristics of the "cooperation" word vector of the registration needs vector are close to the characteristics of the "help" word vector of the inquiry needs vector. The characteristics of the "want" word vector of the registration needs vector are close to the characteristics of the "want" word vector of the inquiry needs vector. Therefore, the registration needs vector related to the first line is close to the inquiry needs vector of "I want you to help develop a quiet vacuum cleaner" as a whole, and shows a high degree of similarity. In particular, it is effective in that the characteristics of the word vector of "development" and the characteristics of the word vector of "cooperation", which are not included in the needs content vector of "I want to increase the application fields of silent motors", are supplemented by the needs classification vector. The sex is increasing.

FIG. 11 is a functional block diagram of the server 100.
Each component of the server 100 is stored in a storage device and hardware including a CPU (Central Processing Unit), a computing unit such as various coprocessors, a storage device such as a memory and a storage device, and a wired or wireless communication line connecting them. It is realized by software that supplies processing instructions to the arithmetic unit. A computer program may be composed of a device driver, an operating system, various application programs located on the upper layers thereof, and a library that provides common functions to these programs. Each of the illustrated blocks shows a block for each function, not a configuration for each hardware.

The server 100 includes a data storage unit 110, a data processing unit 150, and a communication unit 190. The communication unit 190 is in charge of communication processing via the network. The data storage unit 110 stores various data. The data processing unit 150 executes various processes based on the data acquired by the communication unit 190 and the data stored in the data storage unit 110. The data processing unit 150 also functions as an interface between the communication unit 190 and the data storage unit 110.

The communication unit 190 includes a transmission unit 180 for transmitting data and a reception unit 170 for receiving data.
The transmission unit 180 includes a screen data transmission unit 182 that transmits various screen data. The receiving unit 170 includes a registration needs data receiving unit 172 that receives registration needs data and an inquiry needs data receiving unit 174 that receives inquiry needs data.

The data processing unit 150 includes a model generation unit 152, a registration needs vector calculation unit 154, a screen data generation unit 160, and a search unit 162.
The model generation unit 152 executes the distributed representation model generation process. The registration needs vector calculation unit 154 calculates the registration needs vector. The screen data generation unit 160 generates various screen data. The search unit 162 searches for registration needs that meet the inquiry needs.

The data storage unit 110 includes a reference company database 120, a distributed representation model storage unit 130, and a needs database 140. The reference company database 120 stores the reference company table shown in FIG. The distributed representation model storage unit 130 stores the word table shown in FIG. The needs database 140 has a registration needs table storage unit 142, a needs division table storage unit 144, a needs division vector storage unit 146, and a registration needs vector storage unit 148. The registration needs table storage unit 142 stores the registration needs table (see FIG. 5). The needs classification table storage unit 144 stores the needs classification table (see FIG. 6). The needs classification vector storage unit 146 has been described above in relation to FIG. The registration needs vector storage unit 148 has been described above in connection with FIG.

FIG. 12 is a flowchart showing the process of the distributed representation model generation process.
The model generation unit 152 generates a corpus from the reference company table (FIG. 2) (S10). A corpus is a large-scale data that structures sentences in natural language. The corpus is held in the data storage unit 110. Specifically, the model generation unit 152 performs morphological analysis for each sentence extracted from the business content of each company and converts it into a word-separated form. A word-separator is a character string that separates words with a margin. The data in the word-separated format is included in the corpus.

The model generation unit 152 generates a word vector using a corpus (S12). Specifically, the model generation unit 152 obtains a word vector for each word included in the corpus by learning processing of Word2vec, for example. The generated word vector is stored in the distributed representation model storage unit 130 in the form of a word table (FIG. 3).

FIG. 13 is a flowchart showing the process of registration needs addition processing.
When the registration needs data receiving unit 172 receives the registration needs data from the user terminal 200 (S20), the registration needs data receiving unit 172 updates the registration needs table (S22). Specifically, the registration needs data receiving unit 172 sets a new record in the registration needs table, and sets the registered company name, the registration needs content, and the needs classification ID included in the registration needs data in the record.

The registration needs vector calculation unit 154 calculates the sentence vector of each sentence included in the registration needs content (S24). A sentence vector is generated from the word vector of the words contained in the sentence. The generation method is, for example, averaging. That is, the sentence vector is the average vector of the word vectors.

The registration needs vector calculation unit 154 calculates the average of the sentence vectors of each sentence included in the registration needs content, and uses the calculated average vector as the needs content vector (S26). The registered needs vector calculation unit 154 may generate a needs content vector from the sentence vector by a method other than averaging.

As described above, the registration needs vector calculation unit 154 calculates the registration needs vector by the load average of the needs content vector and the needs classification vector (S28). The registration needs vector is stored in the registration needs vector storage unit 148 in association with the registration needs ID. A simple average may be used instead of the weighted average.

FIG. 14 is a flowchart showing the process of the needs matching process.
When the inquiry needs data receiving unit 174 receives the inquiry needs data from the server 100 (S30), the search unit 162 calculates the inquiry needs vector (S32). Specifically, when the inquiry needs are phrases, clauses, or sentences, words are extracted by morphological analysis. If multiple words are separated by spaces or commas in the inquiry needs content, those words are used. A query needs vector is generated from the word vector of the words included in the query needs content. The generation method is, for example, averaging. That is, the query needs vector is the average vector of the word vectors.

The search unit 162 searches for a registration needs vector similar to the query needs vector (S34). Specifically, the search unit 162 calculates the degree of similarity between each registration need vector stored in the registration needs vector storage unit 148 and the inquiry needs vector. Then, the search unit 162 extracts the registration needs content, the needs classification ID, and the registered company name corresponding to the registration needs vector having a similarity equal to or higher than the reference value.

The screen data generation unit 160 generates a matching result screen that displays a list of the extracted registration needs content, needs classification ID, and registered company name, and the screen data transmission unit 182 uses the generated matching result screen data as a user terminal. It is transmitted to 200 (S36). At this time, the rows of registration needs may be sorted in descending order of similarity. Also, the degree of similarity for each registration need may be displayed.

In the present embodiment, the registration needs vector calculation unit 154 calculates the needs content vector from the word vector representing each word included in the content sentence of each registration need, and further, the word vector representing each word included in the title of the needs category. An example is shown in which the needs classification vector is calculated from the above, and each registered needs vector is calculated by a weighted average in which the ratio of the weight of the needs content vector and the weight of the needs classification vector is constant.

[Modification example]
As a modification, the registration needs vector calculation unit 154 sets the first weight of the word vector representing each word included in the content sentence of each registration need to be constant, and sets the first weight of the word vector representing each word included in the title of the needs category to be constant. 2 Each registration need vector may be calculated by a weighted average with a constant weight.

For example, if the first weight is 7 and the second weight is 3, the value of each element in the registration needs vector of the registration needs ID: N0300 in FIG. 5 is the value of the word vector of "quiet", "motor". Multiply the value of the word vector of "", the value of the word vector of "application", the value of the word vector of "field", the value of the word vector of "increase" and the value of the word vector of "want to" by 7, respectively, and "research". The value of the word vector of "development", the value of the word vector of "development", and the value of the word vector of "cooperation" are each multiplied by 3, and the sum of them is divided by 51 (= 7 × 6 + 3 × 3). NS. A simple average may be used instead of the weighted average.

Further, after performing needs matching by inputting inquiry needs, needs matching may be performed again according to the inquiry needs. That way, you can also check compatibility with registration needs that were added later. For example, when a new registration need is added, the similarity with the matching needs received so far may be verified, or the similarity with the past matching needs may be periodically verified.

Therefore, when the registration needs data receiving unit 172 receives the content sentence and the needs classification regarding the new registration needs, the registered needs data receiving unit 172 adds the received registration needs to the needs database 140. The registration needs vector 154 calculates a registration needs vector representing the added registration needs. The search unit 162 determines whether or not the registration needs vector of the added registration needs approximates the inquiry needs vector received in the past. When the registration needs vector of the added registration needs approximates the query needs vector, the notification unit (not shown) sends the information of the added registration needs to an email addressed to the email address of the user who sent the query needs vector. And send. Alternatively, you may only email to the effect that registration needs that meet past inquiry needs have been added. Further, the user who receives this e-mail may be able to log in from the user terminal 200b and check the information of the added registration needs.

In vector averaging, an evaluation index such as TF-IDF (Term Frequency-Inverse Document Frequency) may be used to weight each original word vector, phrase vector, sentence vector, or the like. ..

Further, the needs classification may be received on the inquiry needs reception screen (FIG. 9), and the needs classification ID may be included in the inquiry needs data. The search unit 162 may extract the registration needs that match the received needs category from the registration needs table (FIG. 5), and match the extracted registration needs with the opposite needs. For example, the opposite need for "I want to buy XX" is "I want to sell XX". That is, the search unit 162 replaces "want to buy" included in the inquiry needs content with the opposite word "want to sell", calculates the inquiry needs vector according to the inquiry needs content, and registers similar to the inquiry needs vector. Try to search for the needs vector.

In the embodiment, an example of obtaining the registration needs vector and the inquiry needs vector by the arithmetic mean is shown, but it may be calculated by another method. For example, evaluations and achievements may be accepted as feedback for matching, and the final registration needs vector and / and final inquiry needs vector may be calculated by machine learning using the evaluations and achievements as teacher data. For example, the evaluation value may be increased when the other party can be contacted or a business negotiation is concluded. The number of times a person can be contacted or the number of times a business negotiation is concluded may be used as the actual value.

It may be possible to specify the needs classification when inputting the inquiry needs. Further, the needs category may be defined in advance by associating the opposite needs with each other. At the time of needs matching, a query needs vector may be generated based on the content of the query needs and the opposite needs of the query needs classification. Then, the degree of similarity with the registration needs vector may be calculated using the inquiry needs vector.

The present invention is not limited to the above-described embodiment or modification, and the components can be modified and embodied within a range that does not deviate from the gist. Various inventions may be formed by appropriately combining a plurality of components disclosed in the above embodiments and modifications. In addition, some components may be deleted from all the components shown in the above embodiments and modifications.

100 server, 110 data storage, 120 reference company database, 130 distributed representation model storage, 140 needs database, 142 registration needs table storage, 144 needs classification table storage, 146 needs classification vector storage, 148 registration needs vector storage Unit, 150 data processing unit, 152 model generation unit, 154 registration needs vector calculation unit, 160 screen data generation unit, 162 search unit, 170 reception unit, 172 registration needs data reception unit, 174 inquiry needs data reception unit, 180 transmission unit , 182 screen data transmitter, 190 communication unit, 200 user terminal, 300 display area, 302 input area, 304 selection area, 306 registration button, 308 cancel button, 310 input area, 312 execute button, 314 cancel button, 320 display area 322 display area, 324 display area, 326 detail button

Claims (5)

Regarding the database that manages the content sentences of a plurality of registration needs in association with the needs category, a vector representing each word included in the content sentence of each registration need and a vector representing each word included in the title of the needs category. A calculation unit that calculates a registration needs vector that represents the registration needs based on
The first reception department that accepts the content of inquiry needs and
A search unit that calculates a query needs vector representing the query needs based on a vector representing each word included in the content sentence of the query needs, and searches for a registration needs vector similar to the calculated query needs vector. ,
A needs matching device comprising: at least an output unit for outputting the content sentence of the registration needs represented by the searched registration needs vector. A second reception department that accepts content sentences and needs classifications related to new registration needs and adds the received registration needs to the database.
The calculation unit calculates a registration needs vector representing the added registration needs, and calculates the registration needs vector.
The search unit determines whether or not the registration needs vector of the added registration needs approximates the query needs vector.
The output unit is characterized in that, when the registration needs vector of the added registration needs approximates the inquiry needs vector, at least the content sentence of the added registration needs is output. The described needs matching device. The calculation unit calculates a needs content vector representing the content sentence from the vector representing each word included in the content sentence of each registration need, and represents each word included in the title of the needs category. The needs according to claim 1 or 2, wherein a needs classification vector representing the needs classification is calculated from the vector, and the registration needs vector is calculated based on the needs content vector and the needs classification vector. Matching device. The calculation unit is based on the vector representing each word included in the content sentence of each registration need and the vector representing each word included in the title of the needs category. The needs matching device according to claim 1 or 2, characterized in that Regarding the database that manages the content sentences of a plurality of registration needs in association with the needs category, a vector representing each word included in the content sentence of each registration need and a vector representing each word included in the title of the needs category. A function to calculate the registration needs vector representing the registration needs based on
A function to accept the content of inquiry needs and
A function of calculating a query needs vector representing the query needs based on a vector representing each word included in the content sentence of the query needs, and searching for a registration needs vector similar to the calculated query needs vector.
A program characterized in that at least the function of outputting at least the content sentence of the registration needs represented by the searched registration needs vector is exhibited in the information processing device.

Images