JP2021036407A - Search support system, learning system, search support method and program - Google Patents

Abstract

To provide a search support system, a search support method and a program that can easily select a document, in which content desired by a searcher is written, even when there are a lot of documents in a search result.SOLUTION: An estimation server 50 of a search support system comprises: a task information acquisition unit 51 that acquires task information associated with a task of a searching user; a search result acquisition unit 52 that acquires a search result indicating a document which is searched using a search keyword used for search by the searching user; a second learned model storage unit 55 that stores information indicating a second learned model which is generated by causing a learning model to learn task information and a learning dataset associating a selection result indicating a document selected by the searching user on the basis of the search result; and a selected document estimation unit 53 that estimates a selected document selected by a target user using the task information, and the search result and the second learned model.SELECTED DRAWING: Figure 5

Description

The present invention relates to a search support system, a learning system, a search support method, and a program.

There is a technology to search documents based on search keywords. For example, Patent Document 1 discloses a technique of expanding an input keyword to a phrase indicating a search purpose and displaying it as a candidate for a search keyword. As a result, the searcher can perform a search by designating a search keyword expressing the search purpose, and can search for a document that matches the purpose without increasing the number of searches.

JP-A-2007-133688

However, even if the document matches the search purpose, it is not always the document desired by the searcher. This is because even if the search purpose is the same, the document desired by the searcher differs depending on the position of the searcher. For example, when the phrase indicating the purpose of the search is "car insurance", it is desirable that a general user who wants to take out car insurance can search a document that describes the location of an insurance company that handles car insurance. On the other hand, it is desirable for a person in charge of automobile insurance at an insurance company, etc. to be able to search for documents that describe the requirements for insurance enrollment procedures, specific procedure details, and whether or not there is a reporting obligation.

In particular, companies that handle various procedures in each industry, such as insurance and finance, have a large number of regulatory documents, administrative standards, inspection standards, notices, forms, manuals, internal control documents, etc. for each procedure. Document exists. Therefore, when performing a procedure, if the procedure name or the like is specified for the search purpose and searched, a large number of documents will be applicable. In order to select a document containing the desired content from such a large number of documents, there is a problem that all the documents must be read in detail, which requires a great deal of labor.

The present invention has been made in view of such a situation, and even when a large number of documents are present in the search results, it is possible to easily select a document in which the content desired by the searcher is described. It provides a search support system, a learning system, a search support method, and a program.

In order to solve the above-mentioned problems of the present invention, the present invention has been first learned by learning the degree of similarity of each of the search target documents by using the feature amount of the search target document which is the search target document. Using the first trained model generation unit that generates a model, the business information acquisition unit that acquires business information related to the business of the search user, the search keyword used by the search user for the search, and the first trained model. A search result acquisition unit that acquires search results indicating a document searched from the search target document, a selection record acquisition unit that acquires a selection record indicating a document selected by the search user based on the search result, and the above. The second learned that the relationship between the business information and the search result and the selected result is learned by training the learning model with the learning data set in which the business information and the search result are associated with the selected result. The second trained model generation unit that generates a model, the business information of the target user to be estimated, the search result according to the search keyword used by the target user for the search, and the second trained model. It is a search support system including a selected document estimation unit that estimates a selected document selected by the target user from the search result.

Further, the present invention further includes a display control unit that controls the order in which documents corresponding to the search results are displayed according to the estimation results estimated by the selected document estimation unit in the above-mentioned search support system.

Further, the present invention uses the feature quantity of the search target document, which is the search target document, to generate a first trained model in which the degree of similarity of each of the search target documents is learned. The department, the business information acquisition department that acquires business information related to the business of the search user, the search keyword used by the search user for the search, and the document searched from the search target document using the first learned model. A search result acquisition unit that acquires the indicated search result, a selection record acquisition unit that acquires a selection record indicating a document selected by the search user based on the search result, the business information, the search result, and the selection. A second trained model generation that generates a second trained model in which the relationship between the business information and the search result and the selected performance is learned by training the training model with a training data set associated with the achievements. It is a learning system that includes a department.

Further, in the present invention, the first trained model generation unit learns the degree of similarity of each of the search target documents by using the feature amount of the search target document which is the search target document. Is generated, the business information acquisition unit acquires business information related to the business of the search user, and the search result acquisition unit uses the search keyword used for the search by the search user and the first learned model to perform the search. The search result indicating the document searched from the target document is acquired, and the selection record acquisition unit acquires the selection record indicating the document selected by the search user based on the search result, and the second trained model generation unit. However, the relationship between the business information and the search result and the selected result was learned by training the learning model with a learning data set in which the business information and the search result are associated with the selected result. 2 A trained model is generated, the business information acquisition unit acquires the business information of the target user to be estimated, and the selected document estimation unit obtains the business information of the target user to be estimated and the target. It is a search support method that estimates a selected document selected by the target user from the search result by using the search result according to the search keyword used by the user for the search and the second trained model. ..

Further, the present invention is a program for operating the computer as the learning system described above, and is a program for operating the computer as each part included in the learning system.

According to the present invention, even when a large number of documents are present in the search results, it is possible to easily select a document in which the content desired by the searcher is described.

It is a block diagram which shows the example of the structure of the search support system 1 which concerns on embodiment. It is a block diagram which shows the example of the structure of the 1st learning server 20 which concerns on embodiment. It is a block diagram which shows the example of the structure of the search server 30 which concerns on embodiment. It is a block diagram which shows the example of the structure of the 2nd learning server 40 which concerns on embodiment. It is a block diagram which shows the example of the structure of the estimation server 50 which concerns on embodiment. It is a figure which shows the example of the structure of the business information stored in the business information storage unit 16 which concerns on embodiment. It is a figure which shows the example of the structure of the information stored in the feature quantity information storage unit 24 which concerns on embodiment. It is a sequence diagram which shows the flow of the process performed by the search support system 1 which concerns on embodiment. It is a sequence diagram which shows the flow of the process performed by the search support system 1 which concerns on embodiment.

Hereinafter, embodiments of the invention will be described with reference to the drawings.

The search support system 1 is a system that supports the search by estimating whether or not the searched document is a document desired by the searcher. The search support system 1 is installed in a company that handles various procedures in each industry, such as an insurance industry and a financial industry.

The search support system 1 is a database in which regulatory documents, office standards, inspection standards, notices, format collections, manual collections, internal control documents, etc. related to various procedures are used as documents to be searched (search target documents). It is premised that it has (corresponding to the search target document information storage unit 35 described later).

In addition, the search support system 1 has two processing phases, a learning stage and an execution stage.
In the learning stage, the search support system 1 performs learning for estimating whether or not the document is desired by the searcher, and the trained model used for the estimation (the first trained model and the second trained model, which will be described later). ) Is generated.
In the execution stage, the search support system 1 uses the learned model generated in the learning stage to perform a process of estimating whether or not the searched document is a document desired by the searcher.

By performing the estimation process at the execution stage in this way, for example, among the searched documents, the document estimated to be desired by the searcher can be preferentially displayed. This makes it easier for the searcher to select the document he / she desires, as compared to the case where the searched document is displayed mechanically. Therefore, it is possible to reduce the labor required for the searcher to search.

In the following, a case where processing in the execution stage is performed after passing through the learning stage will be described as an example, but the present invention is not limited to this. The learning stage and the execution stage may be processed simultaneously, and the information obtained in the execution stage may be used for the processing in the learning stage.

FIG. 1 is a block diagram showing an example of the configuration of the search support system 1 according to the embodiment. The search support system 1 includes, for example, a search terminal 10, a first learning server 20, a search server 30, a second learning server 40, and an estimation server 50. In the search support system 1, the components of the search support system 1 (search terminal 10, first learning server 20, search server 30, second learning server 40, and estimation server 50) are via a communication network NW. , Connected to communicate with each other.

First, the learning stage will be described.
The search terminal 10 is a personal computer, a tablet terminal, or the like used for the search.
In the search terminal 10, the searcher inputs business information and search keywords related to the searcher's business. Here, the searcher in the learning stage is an example of a “search user”.

The search terminal 10 outputs business information related to the searcher's business to the second learning server 40. The search terminal 10 outputs the search keyword input by the searcher to the search server 30. The search terminal 10 acquires search results according to the search keyword from the search server 30. For example, the search result includes a plurality of documents among the documents to be searched.

The search terminal 10 displays the search result.
In the learning stage, the search result display is mechanically displayed according to a predetermined algorithm. The predetermined algorithm here may be arbitrary, but for example, the titles of the documents are displayed in alphabetical order and in the order of AIUEO, or the documents are displayed in the order notified by the search server 30.
The searcher selects a desired document from the displayed search results. In the search terminal 10, a selection record indicating a document selected by the searcher (that is, a document desired by the searcher) is input. The search terminal 10 outputs the selection result to the second learning server 40.

The first learning server 20 is a personal computer, a server device, or the like that generates a first learned model in which the degree of similarity between the search target documents is learned based on the feature quantities of the individual search target documents. The first trained model is a model that outputs information indicating a document similar to the predetermined document based on the learned content when information indicating a predetermined document included in the search target document is input. .. The first learning server 20 outputs the information indicating the generated first trained model to the search server 30.

The search server 30 is a personal computer, a server device, or the like that searches for a search target document using a search keyword and a first trained model. The search server 30 outputs information indicating a document searched from the search target document to the search terminal 10 as a search result.

The second learning server 40 is a personal computer, a server device, or the like that generates a second trained model. The second trained model is a model in which the relationship between the searcher's business information and search results and the document selected by the searcher (selection record) is learned based on the results of searches performed in the past. .. The second learning server 40 outputs the information indicating the generated second trained model to the estimation server 50.

The estimation server 50 estimates a selected document indicating a document selected (probably) by the searcher by using the business information of the searcher, the search result of the search by the searcher, and the second trained model. Personal computers, server devices, etc.

Next, the execution stage will be described.
Similar to the learning stage, the search terminal 10 inputs business information and search keywords related to the searcher's business. The search terminal 10 outputs the search keyword input by the searcher to the search server 30. The search terminal 10 acquires search results according to the search keyword from the search server 30. Here, the searcher in the execution stage is a person to be estimated and is an example of a “target user”.

At the execution stage, the search terminal 10 outputs business information related to the searcher's business and search results according to the search keyword to the estimation server 50.

The estimation server 50 estimates the selected document by using the searcher's business information and the search result acquired from the search terminal 10 and the second trained model.
For example, consider the case where the following documents are extracted as search results when the wording "insurance procedure" is entered in the search keyword by the searcher.
1) Regulation document A that describes the requirements for taking out insurance
2) Business standard B that describes the paperwork for the application form for insurance
3) Inspection Criteria C that describes how to inspect the contents of the application form for taking out insurance
4) Notice D to inform you of the revision of the application form for taking out insurance
5) Form collection E that describes the form collection of the application form for taking out insurance

For example, when the estimation server 50 indicates a business (for example, a sales position) for confirming the requirements for taking out insurance to the customer as the business information of the searcher, the searcher sets the regulation document A. It is highly likely that other documents (business standard B, inspection standard C, notice D, and form collection E) will be selected.
Alternatively, when the estimation server 50 indicates the business of managing the application form for taking out insurance (for example, clerical work) as the business information of the searcher, the searcher sends a notification D and a form collection E. It is presumed that it is highly likely that other documents (Regulated Document A, Administrative Criteria B, and Inspection Criteria C) will be selected.
As described above, the estimation server 50 may use information indicating the possibility that the document is selected for each of the plurality of documents shown as the search result as the estimation result. Alternatively, the estimation server 50 may use binary information as to whether or not the document is selected in each of the plurality of documents shown as the search result as the estimation result.

The estimation server 50 outputs the estimation result to the search terminal 10.
The search terminal 10 acquires the estimation result and displays the search result based on the acquired estimation result. For example, the search terminal 10 displays the search results in an order according to the estimation results based on the search results and the estimation results.
The searcher selects a desired document from the displayed search results.
Here, when the information obtained in the execution stage is used for the processing in the learning stage, the search terminal 10 is provided with a selection record indicating a document selected by the searcher (that is, a document desired by the searcher). , Output to the second learning server 40.

As shown in FIG. 1, the search terminal 10 includes, for example, an input unit 11, a search result acquisition unit 12, an estimation result acquisition unit 13, a display unit 14, a display control unit 15, and a business information storage unit 16. To be equipped.

The input unit 11 connects to an input device such as a keyword or a mouse, and acquires information input via the input device. The information acquired by the input unit 11 is, for example, business information, a search keyword, and information indicating a selection record, which is input by a searcher.
The search result acquisition unit 12 acquires search results according to the search keyword. The search result acquisition unit 12 outputs the search keyword input to the input unit 11 to the search server 30, and as a response, acquires the search result according to the search keyword from the search server 30.

The estimation result acquisition unit 13 acquires the business information of the searcher and the estimation result estimated based on the search result. The estimation result acquisition unit 13 outputs the business information of the searcher input to the input unit 11 to the estimation server 50. The estimation result acquisition unit 13 outputs the search results according to the search keyword of the searcher acquired by the search result acquisition unit 12 to the estimation server 50. As a response, the estimation result acquisition unit 13 acquires the business information of the searcher and the estimation result estimated based on the search result from the estimation server 50.

The display unit 14 displays the search result under the control of the display control unit 15.
The display control unit 15 controls the display of the search result by the display unit 14 according to the estimation result. The display control unit 15 acquires search results according to the search keyword of the searcher from the search result acquisition unit 12. The display control unit 15 acquires the business information of the searcher and the estimation result estimated based on the search result from the estimation result acquisition unit 13. Based on the search result and the estimation result, the display control unit 15 preferentially displays, for example, a document that is likely to be selected by the searcher on the display unit 14.

The business information storage unit 16 stores business information. The business information is information about the business of a searcher who uses the search support system 1, and is, for example, a user ID that uniquely identifies a user and a business attribute that indicates an attribute related to the business of the user specified by the user ID. Is the associated information (see FIG. 6).

FIG. 2 is a block diagram showing an example of the configuration of the first learning server 20 according to the embodiment.
As shown in FIG. 2, the first learning server 20 includes, for example, a document information acquisition unit 21, a feature amount extraction unit 22, a first learned model generation unit 23, and a feature amount information storage unit 24.

The document information acquisition unit 21 acquires the document information of the search target document (hereinafter, referred to as search target document information, or simply document information, etc.). Document information is information related to the content described in the document to be searched, for example, the content itself described in the document, the title of the document, the table of contents, the content described in each chapter, the frequency of repeating the described words, and the like. This is information indicating. In addition, when the search target document includes the time when the document was created, the creator, ISBN (International Standard Book Number), etc., such information may be included as the document information. ..
The document information acquisition unit 21 acquires the document information by inputting the document information via, for example, an input unit of the search terminal 10 or an input unit of the first learning server 20 (not shown).

The feature amount extraction unit 22 extracts the feature amount of the search target document by performing natural language processing on the sentence (document) included in the search target document based on the document information. The natural language processing here is a process including morphological analysis, parsing, semantic analysis, or context analysis, and is, for example, Doc2Vec (Document to Vector). The feature quantity of a document is a numerical value (vector expression) of the features in the document. A feature of a document is a feature relating to the content described in the document, such as whether or not a particular wording tends to be repeated in the document. In this case, the numerical value of the characteristics of the document is, for example, the number of times the specific wording is repeated in the document. The feature amount extraction unit 22 extracts the feature amount for any feature in each of the search target documents.

The first trained model generation unit 23 learns the degree of similarity of the search target documents by executing machine learning using the feature quantities of the search target documents extracted by the feature quantity extraction unit 22. Generate a completed model.

The first trained model calculates the degree of similarity between documents based on, for example, the features of the documents to be searched. The degree of similarity here is a degree indicating the closeness (correlation amount) of values obtained by quantifying predetermined features in each document. For example, a high-dimensional vector space expressed in a distributed manner based on the features of documents. Is the distance in. The first trained model determines that the documents are similar to each other when the distance between the documents in the vector space is smaller than a predetermined threshold value. On the other hand, the first trained model determines that the documents are not similar to each other when the distance between the documents in the vector space is larger than a predetermined threshold value.

The first trained model generation unit 23 outputs the information indicating the generated first trained model to the search server 30.
The feature amount information storage unit 24 stores the feature amount information. The feature amount information is information about the feature amount extracted by the feature amount extraction unit 22, and is, for example, extracted from a document ID that uniquely identifies a search target document and a search target document specified by the document ID. This is information associated with the feature amount for each feature (see FIG. 7).

FIG. 3 is a block diagram showing an example of the configuration of the search server 30 according to the embodiment.
As shown in FIG. 3, the search server 30 includes, for example, a search keyword acquisition unit 31, a corresponding document extraction unit 32, a similar document extraction unit 33, a search result output unit 34, and a search target document information storage unit 35. , A first trained model storage unit 36 is provided.

The search keyword acquisition unit 31 acquires the search keyword input by the searcher from the search terminal 10, and outputs the acquired search keyword to the corresponding document extraction unit 32.
The corresponding document extraction unit 32 extracts the corresponding document from the search target document based on the search keyword acquired from the search keyword acquisition unit 31. The relevant document is a document in which a search keyword or a phrase similar to the search keyword is described among the search target documents. The corresponding document extraction unit 32 outputs information indicating the corresponding document to the similar document extraction unit 33 and the search result output unit 34.

The similar document extraction unit 33 extracts a similar document similar to the corresponding document by using the first trained model. The similar document is a document similar to the corresponding document, and the distance from the corresponding document in the vector space is smaller than a predetermined threshold value, which is determined by using the first trained model. The similar document extraction unit 33 outputs information indicating similar documents to the search result output unit 34.
The search result output unit 34 outputs information indicating each document of the corresponding document acquired from the corresponding document extraction unit 32 and the similar document acquired from the similar document extraction unit 33 to the search terminal 10 as a search result.
The search target document information storage unit 35 stores the search target document information. The search target document information is information related to the search target document, and for example, the document ID that uniquely identifies the search target document corresponds to the attribute information such as the title, the number of pages, and the described contents of the search target document. This is the attached information.

The first trained model storage unit 36 stores information indicating the first trained model generated by the first learning server 20. The information indicating the first trained model is, for example, an algorithm for determining whether or not the documents are similar to each other based on the distances in the vector space in a plurality of documents. Alternatively, the information indicating the first trained model may be information indicating the position in the vector space of each search target document, or the documents determined based on the position in the vector space of each search target document are mutually exclusive. It may be a table in which determination results indicating whether or not they are similar are stored.

FIG. 4 is a block diagram showing an example of the configuration of the second learning server 40 according to the embodiment.
As shown in FIG. 4, the second learning server 40 includes, for example, a business information acquisition unit 41, a search result acquisition unit 42, a selection record acquisition unit 43, a learning data set generation unit 44, and a second learning completed unit. A model generation unit 45 is provided.

The business information acquisition unit 41 acquires the business information input by the searcher from the search terminal 10, and outputs the acquired business information to the learning data set generation unit 44.
The search result acquisition unit 42 acquires search results according to the search keyword from the search terminal 10. Alternatively, the search result acquisition unit 42 may acquire search results according to the search keyword from the search server 30. The search result acquisition unit 42 outputs the acquired search result to the learning data set generation unit 44.
The selection record acquisition unit 43 acquires the selection record selected by the searcher from the search terminal 10, and outputs the acquired selection record to the learning data set generation unit 44.
The learning data set generation unit 44 generates a learning data set by combining the business information and the search result linked by one search with the selection result.

The second trained model generation unit 45 generates the second trained model by causing the learning model to perform machine learning using the training data set as a teacher.
The learning model is, for example, CNN (Convolutional Neural Network). The CNN is a model composed of three layers, an input layer, a hidden layer (intermediate layer), and an output layer. Data to be estimated (input data) is input to the input layer. Data (estimation result) indicating the learned result is output from the output layer. The hidden layer performs the core processing of learning. For example, the hidden layer converts the input into a value represented by a function called an activation function (transfer function) and outputs it. For example, the activation function includes, but is not limited to, a rectified linear function, a sigmoid function, a step function, and the like, and any function may be used.

The second trained model generation unit 45 causes the training model to input the input of the training data set. The input data of the learning data set in this embodiment is business information and search results. The learning model outputs the estimation result from the output layer after processing the hidden layer for the input data.

The second trained model generation unit 45 adjusts the processing parameters of the hidden layer of the training model so that the estimation result output from the output layer approaches the output of the training data set. The output of the data set for learning in this embodiment is a selection result.

The second trained model generation unit 45 adjusts the processing parameters of the hidden layer so that the estimation result output from the output layer approaches the output of the training data set for all of the predetermined number of training data sets. Repeat. In this way, the learning model trained so that a value close to the output of the learning data set can be output as an estimation result with respect to the input of the learning data set becomes the second trained model. The second trained model generation unit 45 transmits the information indicating the generated second trained model to the estimation server 50.

FIG. 5 is a block diagram showing an example of the configuration of the estimation server 50 according to the embodiment.
As shown in FIG. 5, the estimation server 50 includes, for example, a business information acquisition unit 51, a search result acquisition unit 52, a selected document estimation unit 53, an estimation result output unit 54, and a second trained model storage unit 55. And.

Since the function of the business information acquisition unit 51 is the same as that of the business information acquisition unit 41, the description thereof will be omitted. The business information acquisition unit 51 outputs the acquired business information to the selection document estimation unit 53.
Since the function of the search result acquisition unit 42 is the same as that of the search result acquisition unit 42, the description thereof will be omitted. The search result acquisition unit 52 outputs the acquired search result to the selected document estimation unit 53.

The selected document estimation unit 53 estimates the selected document to be estimated by using the business information, the search result, and the second trained model. The selected document estimation unit 53 causes the second trained model to input the business information acquired from the business information acquisition unit 51 and the search result acquired from the search result acquisition unit 52. The selection document estimation unit 53 uses the output from the second trained model as the estimation result of estimating the selection document (which will be) selected by the searcher.

The estimation result output unit 54 outputs the estimation result estimated by the selected document estimation unit 53 to the search terminal 10.
The second trained model storage unit 55 stores information indicating the second trained model generated by the second learning server 40. The information indicating the second trained model includes, for example, the configuration of the CNN (the number of hidden layers, the number of units in each layer, the type of activation function, etc.), and the processing parameters of the hidden layers (weighting coefficient between units, etc.). Bias component amount, etc.).

FIG. 6 is a diagram showing an example of the configuration of business information stored in the business information storage unit 16 according to the embodiment. The business information includes, for example, a user ID and business attributes. The user ID is identification information that uniquely identifies the searcher. The business attribute is information indicating an attribute related to the business of the searcher, and is composed of, for example, a department, a job title, and years of enrollment. The departments are classified according to the department to which the searcher belongs. For example, information such as the sales department of the automobile insurance department, the administrative department of the automobile insurance department, and the planning management department of the medical insurance department is shown. Job titles include roles such as general managers and general employees. The number of years of enrollment indicates the number of years of enrollment in the department to which the searcher currently belongs.

FIG. 7 is a diagram showing an example of the configuration of information stored in the feature amount information storage unit 24 according to the embodiment. The feature amount information includes, for example, a document ID, a title, and a feature amount. The document ID is identification information that uniquely identifies the document to be searched. The title is the title of the document to be searched. The feature amount is a feature amount of the search target document, and the feature amount is shown for each feature, for example, feature 1, feature 2, and so on.

Here, the flow of processing performed by the search support system 1 will be described with reference to FIGS. 8 and 9.
FIG. 8 is a sequence diagram showing the flow of processing in the learning stage.
FIG. 9 is a sequence diagram showing a flow of processing in the execution stage.

As shown in FIG. 8, in the learning stage, the first learning server 20 acquires the search target document information (document information of the search target document) (step S10), and the first trained model is based on the acquired document information. Is generated (step S11). The first learning server 20 outputs the information indicating the generated first trained model to the search server 30. The search server 30 stores information indicating the first trained model (step S12).

When the searcher's business information is input, the search terminal 10 acquires the business information (step S13) and outputs the acquired business information to the second learning server 40.
When the search keyword is input by the searcher, the search terminal 10 acquires the search keyword (step S15) and outputs the acquired search keyword to the search server 30.
When the search server 30 acquires a search keyword from the search terminal 10, the search server 30 extracts the corresponding document from the search applicable document based on the acquired search keyword (step S16). The search server 30 extracts a similar document from the search target document based on the corresponding document and the first trained model (step S17). The search server 30 outputs the corresponding document and similar documents as search results to the search terminal 10 (step S18).

When the search terminal 10 acquires the search result from the search server 30, the search terminal 10 displays the search result (step S19). Further, the search terminal 10 outputs the search result acquired from the search server 30 to the second learning server 40. The second learning server 40 acquires the search result (step S20).
When the searcher selects a desired document from the search results, the search terminal 10 acquires the selection record (step S21). The search terminal 10 outputs the acquired selection result to the second learning server 40. The second learning server 40 acquires the selection result (step S22).

The second learning server 40 generates a learning data set based on the business information acquired in step S14, the search result acquired in step S20, and the selection record acquired in step S22 (step S23).
The search support system 1 generates a predetermined number of learning data sets by repeating the processes shown in steps S100 (steps S13 to S23).
The second training server 40 uses the training data set to generate a second trained model (step S24).

As shown in FIG. 9, in the execution stage, the search server 30 stores the first trained model (step S31). Since the process of storing the first trained model is the same process as the process shown in steps S10 to S12 in the learning stage, the description thereof will be omitted.
The estimation server 50 stores the second trained model (step S32). In the process of storing the second trained model, for example, after the process shown in step S24 is executed in the learning stage, the estimation server 50 acquires information indicating the second trained model from the second learning server 40. , It is done by storing the acquired information.

Since the processes shown in steps S33 to S38 are the same as the processes shown in steps S13 to S18 in the learning stage, the description thereof will be omitted.

When the search terminal 10 acquires the search result from the search server 30, the search terminal 10 outputs the search result to the estimation server 50 (step S39).
The estimation server 50 acquires the search result (step S40).
The estimation server 50 estimates the selected document using the business information acquired in step S34, the search result acquired in step S40, and the second learned model stored in step S32 (step S41). The estimation server 50 outputs the estimation result to the search terminal 10.
The search terminal 10 acquires an estimation result from the estimation server 50 (step S42), and controls the display of the search result according to the acquired estimation result (step S43).

As described above, the search support system 1 of the embodiment includes the first trained model generation unit 23, the business information acquisition unit 41, the search result acquisition unit 42, the selection record acquisition unit 43, and the second trained unit. It includes a model generation unit 45, a business information acquisition unit 51, and a selection document estimation unit 53. The first trained model generation unit 23 generates a first trained model in which the degree of similarity of each of the search target documents is learned by using the feature amount of the search target document which is the search target document. The business information acquisition unit 41 acquires business information related to the business of a searcher (an example of a “search user”) in the learning stage. 1 Using the trained model, acquire search results indicating the documents searched from the search target documents. The selection record acquisition unit 43 acquires the selection record indicating the document selected by the searcher in the learning stage based on the search result. The second trained model generation unit 45 causes the learning model to learn the business information and search results related to the searcher's business in the learning stage and the learning data set in which the selection results are associated with the business information and the search results. And a second trained model that learned the relationship with the selection record is generated. The business information acquisition unit 51 acquires the business information of the searcher (an example of the “target user to be estimated”) in the execution stage. The selected document estimation unit 53 uses the business information of the searcher in the execution stage, the search result according to the search keyword used by the searcher in the execution stage for the search, and the second learned model from the search result. , Estimate the selected document selected by the searcher at the execution stage.

As a result, the search support system 1 of the embodiment can present the search result based on the estimation result of estimating the selected document selected by the searcher. Therefore, even when a large number of documents are present in the search result, it is possible to easily select a document in which the content desired by the searcher is described.

Further, the search support system 1 of the embodiment further includes a display control unit 15. The display control unit 15 controls the order in which the documents corresponding to the search results are displayed according to the estimation result estimated by the selection document estimation unit 53. As a result, the search support system 1 of the embodiment can preferentially display documents that are likely to be desired by the searcher, and has the same effect as the above-mentioned effect.

As described above, in the present embodiment, the search target document is assumed to be a regulation document or the like related to various procedures. The search target document is created according to various procedures, and is created as necessary for each department or business that performs the procedure, for example. Alternatively, a department or business is established due to the need to carry out various procedures. For this reason, there is a correlation between business information and selection results. That is, the searcher's business and the selection record are closely related, and it is considered that the searchers belonging to the same business tend to select documents from similar search results in a similar manner. Therefore, it is possible to accurately estimate the document desired by the searcher by performing the estimation using the second trained model in which the relationship between the business information and the search result and the selection result is learned.

The search support system 1 of the embodiment may function as a learning system. In this case, the learning system includes a first learned model generation unit 23, a business information acquisition unit 41, a search result acquisition unit 42, a selection result acquisition unit 43, and a second learned model generation unit 45. As a result, the learning system can generate a second trained model in which the relationship between the business information and the search result and the selection result is learned, and has the same effect as the above-mentioned effect.

All or part of the search support system 1 in the above-described embodiment may be realized by a computer. In that case, the program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read by the computer system and executed. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. The "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, or a storage device such as a hard disk built in a computer system. Further, a "computer-readable recording medium" is a communication line for transmitting a program via a network such as the Internet or a communication line such as a telephone line, and dynamically holds the program for a short period of time. It may also include a program that holds a program for a certain period of time, such as a volatile memory inside a computer system that serves as a server or a client in that case. Further, the above program may be for realizing a part of the above-mentioned functions, and may be further realized for realizing the above-mentioned functions in combination with a program already recorded in the computer system. It may be realized by using a programmable logic device such as FPGA.

Although the embodiments of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs and the like within a range that does not deviate from the gist of the present invention.

1 ... Search support system 10 ... Search terminal 11 ... Input unit 12 ... Search result acquisition unit 13 ... Estimated result acquisition unit 14 ... Display unit 15 ... Display control unit 16 ... Business information storage unit 20 ... First learning server 21 ... Document information Acquisition unit 22 ... Feature amount extraction unit 23 ... First learned model generation unit 24 ... Feature amount information storage unit 30 ... Search server 31 ... Search keyword acquisition unit 32 ... Corresponding document extraction unit 33 ... Similar document extraction unit 34 ... Search result Output unit 35 ... Search target document information storage unit 36 ... First learned model storage unit 40 ... Second learning server 41 ... Business information acquisition unit 42 ... Search result acquisition unit 43 ... Selection record acquisition unit 44 ... Learning data set generation Part 45 ... Second trained model generation unit 50 ... Estimated server 51 ... Business information acquisition unit 52 ... Search result acquisition unit 53 ... Selected document estimation unit 54 ... Estimated result output unit 55 ... Second trained model storage unit

Claims (5)

A first trained model generation unit that generates a first trained model that learns the degree of similarity between the search target documents using the features of the search target document, which is the document to be searched.
The business information acquisition department that acquires business information related to the search user's business,
A search result acquisition unit that acquires search results indicating documents searched from the search target document using the search keyword used by the search user and the first learned model, and a search result acquisition unit.
A selection record acquisition unit that acquires a selection record indicating a document selected by the search user based on the search result, and a selection record acquisition unit.
The second learning that learned the relationship between the business information and the search result and the selected result by letting the learning model learn the data set for learning in which the business information and the search result are associated with the selected result. The second trained model generator that generates the trained model,
Using the business information of the target user to be estimated, the search result corresponding to the search keyword used by the target user in the search, and the second learned model, the target user is obtained from the search result. A selection document estimation unit that estimates the selection document selected by
Search support system equipped with. A display control unit that controls the order in which documents corresponding to the search results are displayed according to the estimation result estimated by the selected document estimation unit is further provided.
The search support system according to claim 1. A first trained model generation unit that generates a first trained model that learns the degree of similarity between the search target documents using the features of the search target document, which is the document to be searched.
The business information acquisition department that acquires business information related to the search user's business,
A search result acquisition unit that acquires search results indicating documents searched from the search target document using the search keyword used by the search user and the first learned model, and a search result acquisition unit.
A selection record acquisition unit that acquires a selection record indicating a document selected by the search user based on the search result, and a selection record acquisition unit.
The second learning that learned the relationship between the business information and the search result and the selected result by letting the learning model learn the data set for learning in which the business information and the search result are associated with the selected result. The second trained model generator that generates the trained model,
Learning system with. The first trained model generation unit generates a first trained model in which the degree of similarity of each of the search target documents is learned by using the feature amount of the search target document which is the search target document.
The business information acquisition department acquires business information related to the search user's business,
The search result acquisition unit acquires the search result indicating the search keyword used for the search by the search user and the document searched from the search target document using the first trained model.
The selection record acquisition unit acquires the selection record indicating the document selected by the search user based on the search result.
The second trained model generation unit causes the learning model to learn the business information, the search result, and the learning data set in which the selected result is associated with the selected result, thereby causing the business information, the search result, and the selected result to be learned. Generate a second trained model that trains the relationship with
The business information acquisition department acquires the business information of the target user to be estimated, and
The selected document estimation unit uses the business information of the target user to be estimated, the search result corresponding to the search keyword used by the target user for the search, and the second learned model to perform the search. From the result, the selected document selected by the target user is estimated.
Search support method. A program for operating a computer as the learning system according to claim 3, wherein the computer functions as each part included in the learning system.

Images