JP7446023B1 - assist server - Google Patents

Abstract

The present invention provides an AI server assist system, an intermediary server, and a server intermediary system that reduce the number of man-hours for inputting data into AI and perform processing on unlearned matters. [Solution] The assist server inputs data to be processed received from an intermediary server connected via a public communication network 4, processes the data to be processed, and outputs processed data to be sent to the intermediary server. If the trained model outputs a response indicating that the trained model has not yet learned the input data to be processed, a search server with a search engine will receive keywords related to matters that the trained model has not yet learned. Access the URL listed in the search results output from the search to obtain the content of the external site, train the trained model with the retrieved content, and re-input the data to be processed into the trained model after training. The trained model acquires the processed data that has been processed again and sends it to the intermediary server. [Selection diagram] Figure 13

Description

The present invention relates to an assist server.

Recently, a technology called AI (Artificial Intelligence) is rapidly developing. Then, a server system equipped with AI, specifically a trained model, and equipped with communication functions (hereinafter referred to as the "AI server") will be made available to the public, creating an environment where developers, etc. can freely use it. provided by.

One type of AI that has been attracting attention recently is a so-called chat-type AI that inputs natural language data and outputs a conversation in natural language based on the input data from a trained model. (For example, Non-Patent Document 1.)

This chat-type AI has been trained using dialogue data from various chat services over the Internet, etc., as training data, and the user can output the desired output through dialogue with the chat-type AI. It is about getting.

Therefore, in order to extract a desired output from a chat-type AI, it is necessary to input information requesting what kind of output is desired, in addition to the prerequisite circumstances, situations, or conditions.

For example, if you are trying to get predicted questions and model answers for the marketing theory final exam of a graduate school you are attending to obtain an MBA (Master of Business Administration) from a chat-type AI, On my computer, I sent out prerequisite information such as what I am aiming for, that I am taking a marketing theory course in graduate school, and the scope of the final exam, as well as a request to "teach me" the expected questions and model answers for the final exam. It must be entered interactively into a terminal such as .

However, this input requires considerable effort. In addition, even if you input dialogue sentences with great effort, the chat-type AI will not be able to output the desired information regarding matters that the trained model of the chat-type AI has not yet learned. At present, it only outputs the following.

Such input man-hours and the need to handle unlearned matters are also occurring in AI other than chat-type AI.

Therefore, there is a need for a server mediation system, a mediation server, an assist system, and an assist server that can reduce the number of man-hours for inputting data into AI and can process unlearned matters.

ChatGPT published at https://openai.com/

The problem to be solved by the present invention is to provide a server mediation system, a mediation server, and an AI server assist system that can reduce the number of man-hours for inputting data into AI and process unlearned matters.

The contents described in the above "Background Art" and "Problem to be Solved by the Invention" indicate the impetus that led to the invention, and do not limit the technical scope of the invention. Nor does it permit a limited interpretation of the technical scope of the present invention (2005 (Gyoke) No. 10042, and Patent Office Examination Guidelines Part II Chapter 2 as of the filing date) (See Section 3.2.1).

The present invention provides a trained model that has learned to input processing target data received from an external server connected via a communication line, process the processing target data, and output processed data to be sent to the external server. If the trained model outputs a response indicating that it has not yet learned the input processing target data, the trained model sends a keyword related to the matter that the trained model has not yet learned to a search server having a search engine. access the URL of the external site listed in the search results output by the search to acquire the content of the external site, cause the trained model to learn the acquired content, and perform the processing after learning. An assist server is provided that includes a control unit that inputs target data into the learned model again and processes it, acquires the processed data that has been processed again by the learned model, and sends it to the external server.

According to the present invention, it is possible to provide a server mediation system, a mediation server, a learned model assist system, and an assist server that can reduce the number of man-hours for inputting data into AI and can process unlearned matters.

FIG. 2 is a diagram showing the configuration of a mediation server and an assist server according to the first embodiment. FIG. 3 is a diagram showing an example of a data structure of a user database. It is a figure showing the data structure of a question database. FIG. 3 is a diagram showing a data structure of an additional information database. FIG. 3 is a diagram showing the data structure of a format database. FIG. 3 is a diagram showing the data structure of an account database. FIG. 3 is a diagram showing the data structure of a session record database. FIG. 2 is a diagram showing the structure of information stored as an AI group. 1 is a flowchart showing operations of a server mediation system, a mediation server, and an AI server assist system according to the first embodiment. FIG. 3 is a diagram showing an example of an answer/request input screen displayed on a user terminal. FIG. 7 is a diagram showing the configuration of a mediation server, a user terminal that communicates with the mediation server, and an assist server that realizes the server mediation system of the second embodiment. It is a diagram showing the data structure of a site database. It is a flowchart showing the operation of a server mediation system, a mediation server, and an AI server assist system of a second embodiment. FIG. 7 is a diagram showing the configuration of a mediation server, a user terminal, an assist server, and a portfolio server according to a third embodiment. FIG. 2 is a diagram showing an example of a data structure of a portfolio database.

Hereinafter, a server mediation system, a mediation server, and an AI server assist system according to an embodiment of the present invention will be described in detail.

(First embodiment)
In the first embodiment, the intermediary server 1 adds additional data necessary to obtain a desired output from the learned model of the assist server 2 to the initial data received from the user terminal 3, and creates pre-transmission data. , and further prepares data to be processed from the pre-transmission data according to the input format specific to the trained model, instructs the assist server 2 to input the data to be processed into the trained model, and transmits the data to assist. The trained model processes the data to be processed and outputs the processed data from the server 2, and outputs the processed data to the user terminal 3 based on the output format specific to the processed data output by the trained model. prepares output data to be used and sends the output data to the user terminal.

The intermediary server 1 mediates the exchange of data between the user terminal 3 and the assist server 2, and the assist server 2 assists the operation of the learned model.

FIG. 1 is a diagram showing the configuration of a mediation server 1, a user terminal 3 that communicates with the mediation server 1, and an assist server 2, which realize the server mediation system of this embodiment.

As shown in FIG. 1, the mediation server 1 includes a mediation server control section 11, a mediation server storage section 12, and a mediation server input/output section/communication section 13.

The mediation server control unit 11 includes a calculation device such as a CPU (Central processing unit). The mediation server control unit 11 sequentially reads and executes software stored in the mediation server storage unit 12.

The mediation server storage unit 12 includes storage devices such as memory, SSD (Solid State Drive), and HDD (Hard Disk Drive).

The mediation server storage unit 12 stores a user DB 121 (hereinafter referred to as DB), an additional information DB 122, and a format DB 123, in addition to the software executed by the mediation server control unit 11.

The user DB 121 stores, for each user, user information that is information regarding user attributes and the like, and round information that indicates a history of a series of information exchanges with the assist server 2.

The additional information DB 122 includes preconditions and information necessary to obtain a desired output from the learned model of the assist server 2, which is added to the initial data that the mediation server control unit 11 receives from the user terminal 3. It stores additional information such as request details indicating whether a particular output is desired, and additional example sentences that are example sentences set for each additional information.

The format DB 123 stores, for each assist server 2, an input format that is a format required for data to be input to a trained model specific to the trained model stored in the assist server 2, and processed data output by the trained model. Contains a specific output format.

The intermediary server input/output unit/communication unit 13 includes input/output devices such as a keyboard, mouse, microphone, and display, and a communication device that communicates with external devices via a network.

The assist server 2 includes an assist server control section 21, an assist server storage section 22, and an assist server input/output section/communication section 23.

The assist server control unit 21 includes a calculation device such as a CPU. The assist server control unit 21 sequentially reads and executes software stored in the assist server storage unit 22.

The assist server storage unit 22 includes a storage device such as a memory, SSD, or HDD.

The assist server storage unit 22 stores an account DB 221, a session record DB 222, and an AI group 223 in addition to the software executed by the assist server control unit 21.

The account DB 221 stores accounts, which are identifiers uniquely assigned to each user of the assist server 2, and various information regarding each account.

The session record DB 222 stores session information indicating a history of a series of information exchanges with the mediation server 1, and chat information such as the contents of chats exchanged within each session.

The AI group 223 stores information regarding each trained model and the trained model for each trained model. The AI group 223 includes a plurality of trained models, an AI operation that is software that determines how to select and use the plurality of trained models, and sequentially combines and operates the selected trained models. It can also include a part and a part.

The assist server input/output unit/communication unit 23 includes input/output devices such as a keyboard, mouse, microphone, and display, and a communication device that communicates with external devices via a network.

The user terminal 3 and the mediation server 1, and the mediation server 1 and the assist server 2, are connected via a public communication network 4 such as the Internet, and communicate with each other.

FIG. 2 is a diagram showing an example of the data structure of the user DB 121. As shown in FIG. 2, the user DB 121 stores a user ID, which is an identifier uniquely assigned to each user, user information, which is information about the user such as name and contact information, and an identifier uniquely assigned to each round. A certain round ID, round information such as an access date and time, a question ID that identifies the question administered to the user, the content of the user's answer to the question, and an assist server ID that identifies the assist server 2 where the round occurred. , and a session ID that identifies the session in which the round occurred.

Here, rounds and sessions will be explained. A session consists of information exchange from when a user uses a user terminal 3 to start exchanging information with the assist server 2 via the intermediary server 1 until the end of a series of information exchanges with the assist server 2. Shows a single unit. Usually, one session occurs regarding one topic, but the topic may change during the session.

A round indicates one exchange of information between the user terminal 3 and the assist server 2 via the intermediary server 1, which occurs in one session. One round begins with the flow of information in the transmitting direction transmitted from the user terminal 3 to the assist server 2 via the intermediary server 1, and begins with the flow of information in the receiving direction received from the assist server 2 via the intermediary server 1. It ends. Therefore, one session may include multiple rounds.

FIG. 3 is a diagram showing the data structure of the question DB. The mediation server storage unit 12 of the mediation server 1 is an arbitrary database that may or may not store a question DB. Therefore, it is not shown in FIG.

The Question DB uses a question ID, which is an identifier uniquely assigned to each question, a non-cognitive skill measurement type that questions the user's behavior assuming situations that would not normally occur to the user, and a non-cognitive skill measurement type that questions the user's behavior based on situations that do not normally occur to the user. Stores question attribute information, which is information about the attributes of questions such as response ability measurement type questions that set a situation and ask the user's response, and the question text.

FIG. 4 is a diagram showing the data structure of the additional information DB 122. As shown in FIG. 4, the additional information DB 122 stores an additional example sentence ID, which is an identifier uniquely assigned to each example sentence that the mediation server 1 adds to the initial data, additional information, and additional example sentences.

FIG. 5 is a diagram showing the data structure of the format DB 123. As shown in FIG. 5, the format DB 123 stores an assist server ID, which is an identifier assigned to each assist server 2, an input format, and an output format.

FIG. 6 is a diagram showing the data structure of the account DB 221. As shown in FIG. 6, the account DB 221 stores an account, which is an identifier uniquely assigned to each user of the assist server 2, identification information for identifying the user, and information for use for educational purposes by schools or for commercial purposes by companies. Stores account attribute information indicating account attributes such as intended use.

FIG. 7 is a diagram showing the data structure of the session record DB 222. As shown in FIG. 7, the session record DB 222 stores the session ID, which is an identifier uniquely assigned to each session, the account that conducted the session, the session start date and time, the session end date and time, and the chats that occurred within the session ( The chat ID, which is an identifier uniquely assigned to each utterance), the chat content for each chat, the fact that the chat was read by the other party (displayed on the user terminal 3, and input to the learned model on the assist server 2) ), which is turned ON when the read flag has been read.

FIG. 8 is a diagram showing the structure of information stored as the AI group 223. As shown in FIG. 8, the AI group 223 is a collection of indexed data such as a database, and includes a trained model ID, which is an identifier uniquely assigned to each trained model, and data input to the trained model. Input format information regarding the input format such as the data type (for example, plain text, XML, binary data, etc.) and data length that the trained model must have, and the data type, data length, etc. of the data output by the trained model. Output format information regarding the output format of , trained model information indicating attributes of the trained model such as image judgment, character judgment, natural language analysis/interpretation, translation, dialogue, etc., and specific functional explanation of the trained model, Stores the body of the trained model. The AI group 223 can also include an AI operation unit.

FIG. 9 is a flowchart showing the operations of the server mediation system, mediation server 1, and AI server assist system of this embodiment. The server mediation system is realized by the cooperation of hardware and software included in the mediation server 1. The AI server assist system is realized through the collaboration of the hardware and software included in the assist server 2, but by cooperating with the server mediation system, advanced use of the trained model included in the assist server 2 is possible. becomes.

As shown in FIG. 9, in step 901, the control unit of the user terminal 3 transmits an instruction to request a question to the mediation server 1 after exchanging various information to start a session with the mediation server 1. .

In step 902, the mediation server control unit 11 transmits a question read randomly from the question DB or selected from questions that have not yet been administered to the user.

In step 903, the control unit of the user terminal 3 transmits the answer input by the user and, if necessary, a request to the intermediary server 1.

Here, steps 901 to 903 may be replaced by an operation in which the intermediary server 1 simply receives data input from the user terminal 3 to start a dialogue, instead of requesting and transmitting a question.

In step 904, the intermediary server control unit 11 based on the response received from the user terminal 3 (or data for starting a dialogue; hereinafter referred to as initial data) and the optionally received request. Then, an assist server 2 having a trained model to be used is selected from one or more assist servers 2. This selection may be made by the mediation server control unit 11 searching a database that stores information regarding the assist server 2 that the mediation server 1 separately stores in the mediation server storage unit 12 .

In step 905, the mediation server control unit 11 searches the user DB based on the user ID received from the user terminal 3, and if the assist server ID, session ID, and round information are stored, these pieces of information are stored. read out. The mediation server control unit 11 specifies which assist server 2, which session, and which round the user wants to continue using the assist server ID, session ID, and round information. In this case, the mediation server control unit 11 selects the specified assist server 2 instead of the assist server 2 selected in step 904.

In step 906, the mediation server control unit 11 adds the read session ID to the initial data.

In step 907, the intermediary server control unit 11 adds additional data such as a request statement to the initial data to generate pre-conversion data. This request text may be a text created by the user and sent from the user terminal 3, or the intermediary server control unit 11 may create the additional information DB based on the additional information selected by the user via the user terminal 3. It may also be an additional example sentence retrieved and read out.

Specific examples are listed below. (See Figure 10.)
(1) Question Q1 administered to the user. Kaishu Katsu is on the Kanrin Maru at sea heading for America. You are a fellow passenger on this ship. What do you suggest Kaishu Katsu do when he goes to America?
(2) User responses (initial data)
A1. I suggest that Kaishu Katsu invite the President of the United States to dinner. Even if the president likes hamburgers, he should have it for dinner. This is to show that we are on equal footing.
(3) I want the request selected by the user (request 3) to be evaluated from an objective perspective.In this case, the intermediary server control unit 11 will collect the following pre-conversion data based on (1) to (3) above. generate.
(Data before conversion: The underlined part is additional data added to the initial data by the mediation server control unit 11.)
I was asked the following questions:
Q1. Kaishu Katsu is on the Kanrin Maru at sea heading for America. You are a fellow passenger on this ship. What do you suggest Kaishu Katsu do when he goes to America?
I answered this question as follows.
A1. I suggest that Kaishu Katsu invite the President of the United States to dinner. Even if the president likes hamburgers, he should have it for dinner. This is to show that we are on equal footing.
Please evaluate A1, which I answered for Q1, from an objective perspective.

In step 908, the intermediary server control unit 11 searches the format DB 123 based on the assist server ID of the assist server 2 selected in steps 904 and 905, and reads out the input format. The mediation server control unit 11 converts the pre-conversion data based on the read input format, and generates processing target data.

In step 909, the mediation server control unit 11 transmits the processing target data to the selected assist server 2.

In step 910, the assist server control unit 21 inputs (inputs) the received data to be processed into the trained model, and causes the trained model to process the data to be processed.

In step 911, the assist server control unit 21 transmits to the intermediary server 1 the processed data output by the learned model processing the processing target data.

In step 912, the mediation server control unit 11 searches the format DB 123 based on the assist server ID and reads out the output format. The mediation server control unit 11 converts the processed data based on the read output format and generates response data to be displayed on the user terminal 3.

In step 913, the mediation server control unit 11 transmits the response data to the user terminal 3.

In step 914, the control unit of the user terminal 3 displays the received response data on the display of the user terminal 3.

Among the above operations, steps 901 to 908 are called round start operations, and steps 912 to 914 are called round end operations.

FIG. 10 is a diagram showing an example of an answer/request input screen 100 displayed on the user terminal 3. When transmitting the question in step 902 of FIG. 9, the mediation server control unit 11 generates data for the answer/request input screen 100, transmits it to the user terminal 3, and displays it.

As shown in FIG. 10, the answer/request input screen 100 includes a question display field 1001 for displaying questions, an answer input field 1002 for the user to enter an answer, and a request selection field 1003 that is a pull-down menu for selecting a request. , a request details field 1004 that is displayed when the request selection field 1003 is clicked, and a send button 1005 that instructs to send input/selected information to the intermediary server 1.

As described above, in the server mediation system, the mediation server, and the AI server assist system of the present embodiment, the mediation server 1 allows the user to use the user terminal An additional information DB 122 stores additional information that is data added to the initial data sent from 3, and an input DB 122 that is a format required for data to be input to a trained model specific to a trained model for each assist server 2. The format and the additional information selected from the additional information DB 122 are read out from the format DB 123 that stores the output format specific to the processed data output by the trained model, and the additional information is added to the initial data as additional data and converted. Generate previous data, read the input format from the format DB 123 based on the selected assist server 2, convert the pre-conversion data based on the input format to generate processing target data, and transfer the processing target data to the assist server. 2, the trained model processes the data to be processed and outputs the processed data from the assist server 2, and sends the data to the format DB 123 for output based on the assist server 2. It includes an intermediary server control unit 11 that reads the format, converts the processed data based on the output format, generates response data to be displayed on the user terminal 3, and transmits the response data to the user terminal 3.

Therefore, according to the present embodiment, there is an effect that it is possible to provide a server mediation system, a mediation server, and an AI server assist system that can reduce the number of man-hours for inputting data to be input to AI.

(Second embodiment)
In the second embodiment, the configuration and operation of the mediation server 1 are similar to those of the mediation server 1 in the first embodiment.

On the other hand, the assist server 2 of the second embodiment further stores a site DB 224 in the assist server storage unit 22.

Then, when the assist server control unit 21 of the second embodiment outputs a response indicating that the trained model has not yet learned the input processing target data, the assist server control unit 21 outputs the site URL (uniform Resource Locator), search by sending keywords related to matters that you have not learned to the read site URL, access the URL of external site 6 listed in the search results output from the search, and view the content of that site. The configuration and operation of the assist server 2 in the first embodiment are different in that the acquired content is learned by a trained model, and after learning, the data to be processed is input again and processed, and the processed data is acquired. The behavior is different.

FIG. 11 is a diagram showing the configuration of the mediation server 1, the user terminal 3 that communicates with the mediation server 1, and the assist server 2, which realize the server mediation system of this embodiment.

The assist server 2 of this embodiment further stores a site DB 224 in the assist server storage unit 22. Further, the assist server 2 further connects to a search server 5 and an external site 6 via a public communication network 4 such as the Internet, and communicates with each other.

Here, the search server 5 includes a search engine that, when a keyword is input, outputs a list and URLs of external sites 6 related to the keyword. Furthermore, when the URL is accessed, the external site 6 transmits the content to the access source.

FIG. 12 is a diagram showing the data structure of the site DB 224. As shown in FIG. 12, the site DB 224 contains a site ID, which is an identifier uniquely assigned to a site, site attribute information such as a field of expertise when the search engine has a unique field, and a site URL indicating the URL of the site. Store.

FIG. 13 is a flowchart showing the operations of the server mediation system, mediation server 1, and AI server assist system of this embodiment. The server mediation system is realized by the cooperation of hardware and software included in the mediation server 1. The AI server assist system is realized by the collaboration of the hardware and software included in the assist server 2, but by cooperating with the server mediation system, it is possible to use the trained model included in the assist server 2 even more advancedly. becomes possible.

Specifically, it was difficult for trained models in the prior art to respond to matters that have not been learned, but according to the present embodiment, the assist server 2 responds to matters that have not yet been learned in a trained model. Using a search server 5 having a search engine, find the URL of an external site 6 that describes the matter, access the external site 6, acquire content related to the matter, and acquire the content into the learned model. By learning the model, it assists the further learning of the trained model and realizes further advanced use of the trained model.

In this embodiment, it is possible to use the contents of all sites around the world that can be connected via the public communication network 4 such as the Internet as the background for learning the trained model, and to further improve the trained model. In addition to improving performance, it is also expected to eliminate the time-consuming and labor-intensive tasks of entering keywords and checking search results.

As shown in FIG. 13, in step 1301, the control unit of the user terminal 3 and the mediation server control unit 11 execute a round start operation.

In step 1302, the mediation server control unit 11 transmits the data to be processed to the assist server 2.

In step 1303, the assist server control unit 21 inputs the processing target data received from the intermediary server 1 into the learned model.

In step 1304, the assist server control unit 21 obtains processed data output by the learned model processing the processing target data.

In step 1305, the assist server control unit 21 determines whether the processed data includes content indicating that the data has not been learned. If the assist server control unit 21 determines that the processed data does not include content indicating that it has not been learned (N in step 1305), the process proceeds to step 1313, and the assist server control unit 21 determines that the processed data does not include content indicating that it has not been learned. (Y in step 1305), the process advances to step 1306.

In step 1306, the assist server control unit 21 extracts unlearned matters, especially nouns, as keywords from the processed data, and searches the site DB 224 for search engines that specialize in the field of the keywords, or other similar searches. If the engine does not find a hit, the URL of the search server 5 having a search engine that can search for general items is read. Then, the assist server control unit 21 accesses the read search server 5 and transmits and inputs the keyword as search information to the search engine.

In step 1307, the search server 5 transmits to the assist server control unit 21 the search result containing the URL of the external site 6 containing information related to the keyword.

In step 1308, the assist server control unit 21 reads the URL of the external site 6 from the search results and accesses the external site 6.

In step 1309, the control unit of the external site 6 transmits the content to the assist server 2.

In step 1310, the assist server control unit 21 generates teacher data with the processing target data as an explanatory variable and the content as an objective variable, and causes the trained model to learn the generated teacher data.

In step 1311, the assist server control unit 21 inputs the processing target data into the learned model again.

In step 1312 , the assist server control unit 21 reacquires the processed data output by the trained model after processing the processing target data, and transmits it to the intermediary server 1 .

In step 1313, the control unit of the user terminal 3 and the mediation server control unit 11 execute a round end operation.

As described above, in the server mediation system, mediation server, and AI server assist system of this embodiment, the assist server 2 further stores the site DB 224 in the assist server storage unit 22, and the assist server control unit 21 performs learning. If the completed model outputs a response indicating that it has not yet learned about the input processing target data, the site URL of the search server 5 is read from the site DB 224, and keywords related to the matters that have not been learned are sent to the read site URL. , access the URL of the external site 6 listed in the search results output by the search, acquire the content of the external site 6, train the learned model with the acquired content, and after learning, process the data. Input it again, process it, and get the processed data.

Therefore, the assist server 2 uses the search server 5 having a search engine to find the URL of the external site 6 that describes the matters that have not yet been learned by the trained model, accesses the external site 6, and accesses the information. There is an effect that it is possible to acquire content related to a matter, have a trained model learn the acquired content, and after learning, input the data to be processed again and have it processed, thereby acquiring processed data.

(Third embodiment)
In the third embodiment, the configuration and operation of the assist server 2 are similar to the configuration and operation of the assist server 2 in the second embodiment.

On the other hand, the mediation server 1 of the third embodiment connects to the portfolio server 8 via the public communication network 4, and the mediation server control unit 11 allows the user to access the portfolio database 821 of the portfolio server 8 in advance from the portfolio server 8. The intermediary server 1 of the first embodiment and the second embodiment has the advantage that the acquired portfolio is acquired along with additional data to the initial data when generating pre-conversion data. The behavior is different from that of

Below, the configuration of the portfolio server 8 will be described first, and then the operation of the intermediary server control unit 11 to acquire a portfolio from the portfolio DB 821 of the portfolio server 8 will be described.

FIG. 14 is a diagram showing the configuration of the mediation server 1 that implements the server mediation system of this embodiment, the user terminal 3 that communicates with the mediation server 1, the assist server 2, and the portfolio server 8.

Since the configurations of the mediation server 1 and the assist server 2 have already been explained, the configuration of the portfolio server 8 will be explained here.

As shown in FIG. 14, the portfolio server 8 includes a portfolio server control section 81, a portfolio server storage section 82, and a portfolio server input/output section/communication section 83.

The portfolio server control unit 81 includes a calculation device such as a CPU. The portfolio server control unit 81 sequentially reads and executes the software stored in the portfolio server storage unit 82.

The portfolio server storage unit 82 includes storage devices such as memory, SSD, and HDD.

The portfolio server storage unit 82 stores a portfolio DB 821 in addition to the software executed by the portfolio server control unit 81.

The portfolio server input/output unit/communication unit 83 includes input/output devices such as a keyboard, mouse, microphone, and display, and a communication device that communicates with external devices via a network.

FIG. 15 is a diagram showing an example of the data structure of the portfolio DB 821. As shown in FIG. 15, the portfolio DB 821 stores a portfolio ID, which is an identifier uniquely assigned to a portfolio, and portfolio data, which is the body of the portfolio, for each user ID.

Next, regarding the operation by which the mediation server control unit 11 acquires a portfolio from the portfolio DB 821 of the portfolio server 8, only the portions that differ from the operation of the mediation server control unit 11 in the first embodiment will be described.

The operation of step 906 in the first embodiment is changed as follows.
"In step 906, the mediation server control unit 11 transmits an instruction to the portfolio server 8 to read portfolio data from the portfolio DB 821 and transmit it to the mediation server 1 based on the user ID received from the user terminal 3. The control unit 11 adds the session ID and the portfolio data received from the portfolio server 8 to the initial data.

As described above, in the server mediation system, mediation server, and AI server assist system of this embodiment, the mediation server 1 connects to the portfolio server 8 via the public communication network 4, and the mediation server control unit 11 connects to the portfolio server 8 via the public communication network 4. A portfolio that the user has previously stored in the portfolio DB 821 of the portfolio server 8 is acquired from the portfolio server 8, and the acquired portfolio is added to the initial data together with additional data when generating pre-conversion data.

Here, we will explain the portfolio. In recent years, in entrance exams such as university entrance exams, or personnel evaluations or personnel selection in companies, it has become important to not only examine academic ability, but also to examine students' participation in volunteer activities in addition to various activities at school during their student days. Attempts are being made to have students submit portfolios that compile their activity records, such as information on their experience winning sports awards and traveling abroad, and to include the portfolios in their evaluations.

Spreadsheet software, word processing software, etc. can be used to generate a portfolio, but recently cloud-based portfolio creation systems have been used to reduce man-hours and to encourage mutual stimulation through contact with other users. In addition to the applicant, several companies are providing systems that enable this.

In this case, this portfolio represents the user's values, direction of interests, or part of his or her personality.

Therefore, according to the present embodiment, the intermediary server 1 adds the user's portfolio data to the initial data to generate pre-conversion data, format-converts the pre-conversion data to generate processing target data, and performs the process. Since the target data is input to the trained model, compared to the case where portfolio data is not added, processed data that is more suited to the user's personality is output.

1 Mediation server 2 Assist server 3 User terminal 4 Public communication network 5 Search server 6 External site 8 Portfolio server 11 Mediation server control section 12 Mediation server storage section 13 Mediation server input/output section/communication section 21 Assist server control section 22 Assist server Storage unit 23 Assist server input/output unit/communication unit 81 Portfolio server control unit 82 Portfolio server storage unit 83 Portfolio server input/output unit/communication unit 100 Answer/request input screen 223 AI group 1001 Question display field 1002 Answer input field 1003 Request selection Column 1004 Request details column 1005 Send button

Claims (4)

a trained model that has learned to input processing target data received from an external server connected via a communication line, process the processing target data, and output processed data to be transmitted to the external server;
When the trained model outputs a response indicating that it has not yet learned regarding the input data to be processed, the trained model transmits keywords related to matters that the trained model has not yet learned to a search server having a search engine. make a search, access the URL of an external site listed in the search results output by the search, acquire the content of the external site, make the learned model learn the acquired content, and after learning, process the data to be processed. a control unit that inputs the data into the trained model again and processes it, acquires the processed data that has been processed again by the trained model, and sends it to the external server;
Assist server with. comprising a plurality of the trained models,
The control unit includes:
Claim 1 further comprising a plurality of trained models and an AI operation unit that determines how to select and use the plurality of trained models and sequentially combines and operates the selected trained models. Assist server described in. The trained model is
Claim 1 or 2, wherein the trained model is trained to input the data to be processed, process the data to be processed, and output the processed data such that a dialogue is established with respect to the data to be processed. The assist server according to claim 2. The external server is
an additional information database that stores additional information that is data to be added to initial data transmitted by a user from a user terminal in order to obtain a desired output from the trained model possessed by the assist server;
For each of the assist servers, an input format that is a format required for data input to the trained model that is specific to the trained model, and an output format that is unique to the processed data output by the trained model. a format database that stores
Read the additional information selected from the additional information database, add the additional information to the initial data as additional data to generate pre-conversion data, and convert the input from the format database based on the selected assist server. read out a format for processing, convert the pre-conversion data based on the input format to generate the data to be processed, send the data to be processed to the assist server, input it to the learned model and process it. , receives from the assist server the processed data that the trained model has processed and outputted the data to be processed, reads the output format from the format database based on the assist server, and converts the output format into the output format. an intermediary server control unit that converts the processed data based on the data, generates response data to be displayed on the user terminal, and transmits the response data to the user terminal;
The assist server according to claim 3, which is an intermediary server comprising:

Images