Artificial intelligence system and program for artificial intelligence system

Abstract

To provide an artificial intelligence system and its program for allowing a user to obtain an answer based on relevant information which is the latest information such as current news in real time.SOLUTION: An artificial intelligence system 100 includes a user terminal 110 and servers 120, 130, and 140. When text is input and transmission operation is performed on a chat screen 112 displayed by a browser 111 of the user terminal, the user terminal transmits text data to the server 120 as question information QT. The server 120 acquires relevant information RT related to the question information from the Internet, executes text data generation of contents with respect to the question information using text generation means 131 which is artificial intelligence means trained in advance on the basis of the relevant information RT and the question information QT, and transmits the obtained generated text data GT to the user terminal 110, and the user terminal 110 displays the generated text data GT on the chat screen 112 of the browser 111.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an artificial intelligence system including a user terminal and a server, in which the user terminal can freely communicate with pre-trained artificial intelligence means of the server, and a program therefor. [Background technology]

[0002] Conventionally, there is known an information processing device that includes a language data receiving unit that receives language data in response to a request from a user terminal, a language data processing unit that processes the language data to generate basic summary information that forms the basis for generating a summary, a transmitting unit that transmits presentation information including the basic summary information to the user terminal, an editing information receiving unit that receives editing information related to the presentation information from the user terminal, and a summary generation unit that provides the editing information to a large-scale language model that receives input of the basic summary information and generates a summary sentence that reflects the editing information (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2023-73095 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the conventional information processing devices described above were limited to generating text summaries directly from basic summary information in some language text format using a large-scale language model, which made it difficult for users to obtain answers based on relevant information, such as current news, in real time.

[0005] Therefore, the present invention solves the problems of the prior art as described above, that is, the object of the present invention is to provide an artificial intelligence system and a program therefor that allow users to obtain answers based on relevant information, such as current news, in real time. [Means for solving the problem]

[0006] The invention of claim 1 is an artificial intelligence system comprising a user terminal and a server, wherein the user terminal is capable of communicating freely with pre-trained artificial intelligence means of the server, wherein the user terminal communicates with the server, the browser of the user terminal displays a talk screen, the talk screen has a talk display area that displays question information and generated text data, which are interactions between the user and the pre-trained artificial intelligence means, and when text is input in the talk display area of ​​the talk screen and a send operation is performed, the user terminal sends the text data to the server as question information, and the server obtains related information related to the received question information from the Internet, or the browser of the user terminal obtains related information related to the question information from the Internet as text data as question information and sends the question information and related information to the server, and the server generates text data of the content of the question information using text generation means, which is pre-trained artificial intelligence means, based on the related information and question information, and sends the generated text data obtained by the execution to the user terminal, and the user terminal displays the received generated text data in the talk display area of ​​the browser's talk screen, thereby solving the above-mentioned problem.

[0007] The invention of claim 2 further solves the above-mentioned problem by configuring the artificial intelligence system as defined in claim 1 such that, in addition to the configuration of the artificial intelligence system defined in claim 1, the talk screen has a real-time internet reference button, and when the real-time internet reference button is operated to switch reference ON, and then text is input and a send operation is performed, the server obtains related information related to the received question information from the Internet, or the browser of the user terminal obtains related information related to the question information from the Internet as text data for the question information, and transmits the question information and related information to the server, and the server generates text data of the content of the question information using a text generation means which is a pre-trained artificial intelligence means based on the related information and the question information; on the other hand, when the real-time internet reference button is not operated and reference is OFF, and text is input and a send operation is performed, the server does not obtain related information related to the received question information from the Internet, or the browser of the user terminal does not obtain related information related to the question information from the Internet as text data for the question information, and transmits the question information to the server, and the server generates text data of the content of the question information using a text generation means which is a pre-trained artificial intelligence means, thereby further solving the above-mentioned problem.

[0008] The invention of claim 3 further solves the above-mentioned problem by being configured such that, in addition to the configuration of the artificial intelligence system described in claim 1 or claim 2, when the server retrieves related information related to the received question information from the Internet, or when the browser of the user terminal retrieves related information related to the question information from the Internet using text data as question information, the server refers to the update date and time information of the related information and retrieves one or more pieces from the top, giving priority to those with more recent update date and time information, or retrieves one or more pieces of information with update date and time information from a predetermined period in the past based on the current date and time.

[0009] The invention according to claim 4 further comprises the configuration of the artificial intelligence system according to claim 3, wherein the server converts the received question information into vector coordinate system question data using vector coordinate conversion means, acquires a plurality of pieces of related information related to the received question information from the Internet, converts the acquired plurality of pieces of related information into vector coordinate system related data using vector coordinate conversion means, searches for vector coordinate system related data that is most closely related to the vector of the vector coordinate system question data, and selects the most closely related vector coordinate system related data or related information corresponding to the most closely related vector coordinate system related data; or the browser of the user terminal converts the question information, using text data as question information, into vector coordinate system question data using vector coordinate conversion means, acquires a plurality of pieces of related information related to the question information from the Internet, converts the acquired plurality of pieces of related information into vector coordinate system related data using vector coordinate conversion means, searches for vector coordinate system related data that is most closely related to the vector of the vector coordinate system question data, and selects the most closely related vector coordinate system related data or related information corresponding to the most closely related vector coordinate system related data. The problem described above is further solved by the following configuration: the server selects related vector coordinate system related data or related information corresponding to multiple top-ranked vector coordinate system related data that are closely related, attaches the selected related information to the question information, and transmits the question information to the server; alternatively, the browser of the user terminal obtains multiple pieces of related information related to the question information from the Internet, and transmits the question information and the multiple pieces of related information to the server; the server converts the question information into vector coordinate system question data using the vector coordinate conversion means, and converts the multiple pieces of related information into vector coordinate system related data using the vector coordinate conversion means, searches for vector coordinate system related data that is closest to the vector of the vector coordinate system question data, and selects the most closely related vector coordinate system related data or related information corresponding to the multiple top-ranked vector coordinate system related data that are closely related; and the server generates text data using the text generation means, which is the pre-trained artificial intelligence means, based on the selected related information and question information, and transmits the generated text data obtained by the execution to the user terminal.

[0010] The invention of claim 5 further solves the above-mentioned problem by being configured such that, in addition to the configuration of the artificial intelligence system described in claim 4, when the server obtains related information related to the received question information from the Internet, or when the browser of the user terminal obtains related information related to the question information from the Internet using text data as question information, related information is obtained within a specified range based on URL information of a website, which is a pre-specified electronic information source, on an administrator terminal that can communicate with the server.

[0011] The invention of claim 6 is a program for an artificial intelligence system in which a user terminal can communicate with pre-trained artificial intelligence means of a server, the program comprising: a talk screen display step in which the user terminal communicates with the server, and the browser of the user terminal displays a talk screen having a talk display area in which question information and generated text data, which are interactions between the user and the pre-trained artificial intelligence means; a related information acquisition step in which, when text is input and a send operation is performed in the talk display area of ​​the talk screen, the user terminal transmits the text data to the server as question information, and the server retrieves related information related to the received question information from the Internet, or the browser of the user terminal retrieves related information related to the question information from the Internet as text data, and transmits the question information and related information to the server; a text generation and transmission step in which the server generates text data of the content of the question information using text generation means, which is pre-trained artificial intelligence means, based on the related information and question information, and transmits the generated text data obtained by the execution to the user terminal; and a generated text display step in which the user terminal displays the received generated text data in the talk display area of ​​the browser's talk screen. [Effects of the Invention]

[0012] The artificial intelligence system of the present invention is equipped with a user terminal and a server, which not only enables the user terminal to communicate with the server's pre-trained artificial intelligence means, but also provides the following unique effects.

[0013] According to the artificial intelligence system of the invention of claim 1, related information is obtained from the Internet in real time and learned using so-called RAG (Retrieval Augmented Generation), which is an external information source for the artificial intelligence system. Based on the question information and related information, generated text data is generated as an answer and displayed on the talk screen, allowing the user to obtain answers based on related information, which is the latest information such as current news, in real time.

[0014] According to the artificial intelligence system of the invention of claim 2, in addition to the effect achieved by the invention of claim 1, it is possible to freely select whether or not to acquire related information in real time. Therefore, when reference is OFF, it does not take time to acquire related information, and emphasis is placed on response. On the other hand, when reference is ON, it takes time to acquire related information, but the latest related information is referenced, and emphasis is placed on the freshness of the information.

[0015] According to the artificial intelligence system of the invention of claim 3, in addition to the effects achieved by the invention of claim 1 or claim 2, the update date and time information is referenced, and relatively old related information on the Internet is not obtained, but relatively new information is obtained, and the content of the generated text data as an answer to the question information does not include relatively old information, but includes relatively new information, so that the user can obtain an answer based on related information that is more reliably the latest information than current news, etc., in real time.

[0016] According to the artificial intelligence system of the invention of claim 4, in addition to the effect of the invention of claim 3, the quality of the answer can be improved because in the artificial intelligence system, only related information that is highly relevant to the question is added to the question information and sent to the text generation means. Furthermore, only related information from a selected small portion of external information sources that is highly relevant to the question is added to the question information and sent to the text generation means, and the amount of data sent to the text generation means is relatively small, which prevents the response time from sending a question to receiving an answer from becoming long.

[0017] According to the artificial intelligence system of the invention of claim 5, in addition to the effects achieved by the invention of claim 4, by having the administrator specify in advance the range of reliable electronic information sources, relevant information can be obtained on the Internet within the range of the pre-specified electronic information sources, so that unreliable electronic information sources can be excluded and RAG learning can be performed in real time within the range of reliable electronic information sources, and the user can obtain answers to question information from reliable electronic information sources based on RAG learning.

[0018] According to the program of the artificial intelligence system of the invention of claim 6, similar to the effect achieved by the invention of claim 1, related information is obtained from the Internet in real time and learned by so-called RAG (Retrieval Augmented Generation), which is an external information source for the artificial intelligence system, and generated text data as an answer based on the question information and related information is generated and displayed on the talk screen, so that the user can obtain answers based on related information, which is the latest information such as current news, in real time. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a diagram showing the concept of an artificial intelligence system according to an embodiment of the present invention; [Figure 2] FIG. 2 is a chart showing an example of the operation of the artificial intelligence system according to the embodiment of the present invention. [Figure 3] FIG. 10 is a diagram showing an example of a talk screen on a browser of a user terminal of an artificial intelligence system according to an embodiment of the present invention. [Figure 4]A figure showing the state when question information is sent on the talk screen of the browser of a user terminal of an artificial intelligence system which is an embodiment of the present invention. [Figure 5] A figure showing how generated text data is displayed as an answer on the talk screen of a browser on a user terminal of an artificial intelligence system that is an embodiment of the present invention. [Figure 6] (A) and (B) are explanatory diagrams showing how the first server selects a predetermined number of pieces of related information related to the received question information on the Internet by referring to update date and time information, and how it selects information from a predetermined period before the current time. [Figure 7] FIG. 10 is an explanatory diagram showing how selected related information is added to question information of an artificial intelligence system according to an embodiment of the present invention and transmitted to a pre-trained artificial intelligence means. [Figure 8] FIG. 1 is an explanatory diagram showing how question information, which is a prompt of the artificial intelligence system according to an embodiment of the present invention, is converted into vector coordinate system question data. [Figure 9] FIG. 10 is an explanatory diagram showing how related information acquired by the artificial intelligence system according to an embodiment of the present invention is converted into vector coordinate system related data. [Figure 10] FIG. 1 is an explanatory diagram showing how vector coordinate system-related data of an artificial intelligence system according to an embodiment of the present invention is stored in a vector index. [Figure 11] 1A and 1B are explanatory diagrams illustrating correlation search between vector coordinate system query data and vector coordinate system related data in an artificial intelligence system according to an embodiment of the present invention. [Figure 12] FIG. 10 is an explanatory diagram showing a first modified example of how selected related information is added to question information of the artificial intelligence system according to an embodiment of the present invention and transmitted to the pre-trained artificial intelligence means. [Figure 13] FIG. 10 is an explanatory diagram showing a second modified example of how selected related information is added to question information of an artificial intelligence system according to an embodiment of the present invention and transmitted to a pre-trained artificial intelligence means. [Figure 14] FIG. 2 is a diagram showing an example of a reference range specification window of a browser on an administrator terminal of an artificial intelligence system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] The artificial intelligence system of the present invention is configured such that a user terminal communicates with a server, the browser of the user terminal displays a talk screen, the talk screen has a talk display area that displays question information and generated text data, which are interactions between the user and the pre-trained artificial intelligence means, and when text is input in the talk display area of ​​the talk screen and a send operation is performed, the user terminal sends the text data to the server as question information, and the server obtains related information related to the received question information from the Internet, or the browser of the user terminal obtains related information related to the question information from the Internet as text data as question information and sends the question information and related information to the server, the server generates text data for the content of the question information using a text generation means, which is the pre-trained artificial intelligence means, based on the related information and question information, and sends the generated text data obtained by this generation to the user terminal, and the user terminal displays the received generated text data in the talk display area of ​​the browser's talk screen, so that the user can obtain answers based on related information, which is the latest information such as current news, in real time.The specific implementation is not limited to this. In addition, the program for the artificial intelligence system of the present invention includes a talk screen display step in which a user terminal communicates with a server, and the browser of the user terminal displays a talk screen having a talk display area that displays question information and generated text data, which are interactions between the user and the pre-trained artificial intelligence means; a related information acquisition step in which, when text is input and a send operation is performed in the talk display area of ​​the talk screen, the user terminal sends the text data to the server as question information, and the server obtains related information related to the received question information from the Internet, or the browser of the user terminal obtains related information related to the question information from the Internet as text data as question information, and sends the question information and related information to the server; a text generation and transmission step in which the server generates text data for the content of the question information using a text generation means, which is the pre-trained artificial intelligence means, based on the related information and question information, and transmits the generated text data obtained by this execution to the user terminal; and a generated text display step in which the user terminal displays the received generated text data in the talk display area of ​​the browser's talk screen, thereby allowing the computer to execute the following steps, so that the user can obtain answers based on related information, which is the latest information such as current news, in real time.

[0021] For example, the user terminal may be a desktop personal computer terminal, a notebook personal computer terminal, a smartphone terminal, a tablet terminal, or the like that sends and receives information, and may be any terminal that can be connected to the server via a communication network including a wide area network such as the Internet, a local network, or a telephone line. The servers may be a single server or multiple servers on a cloud. The text generation means, which is a pre-trained artificial intelligence means, is a text generation means that generates text data and is an interactive text generation means also known as a large-scale language model, such as ChatGPT (Generative Pre-trained Transformer) (hereinafter referred to as ChatGPT), and may be configured to be installed on one server or multiple servers on the cloud. Furthermore, the related information may be any electronic data, such as electronic file information such as document files on the Internet, website information, or database information. [Example]

[0022] An artificial intelligence system 100 according to an embodiment of the present invention will be described below with reference to FIGS. 1 is a diagram showing the concept of an artificial intelligence system 100 according to an embodiment of the present invention, FIG. 2 is a chart showing an example of the operation of the artificial intelligence system 100 according to an embodiment of the present invention, FIG. 3 is a diagram showing an example of a talk screen 112 of a browser 111 of a user terminal 110 of the artificial intelligence system 100 according to an embodiment of the present invention, FIG. 4 is a diagram showing a state when question information QT is sent on the talk screen 112 of the browser 111 of the user terminal 110 of the artificial intelligence system 100 according to an embodiment of the present invention, FIG. 5 is a diagram showing a state when generated text data GT is displayed as an answer on the talk screen 112 of the browser 111 of the user terminal 110 of the artificial intelligence system 100 according to an embodiment of the present invention, FIG. 6(A) is an explanatory diagram showing a state in which the first server 120 selects a predetermined number of top related information RT related to the received question information QT by referring to update date and time information RN when acquiring the received related information RT on the Internet, and FIG. 6(B) is an explanatory diagram showing a state in which the first server 120 selects a predetermined number of top related information RT related to the received question information QT by referring to update date and time information RN when acquiring the received related information RT on the Internet. 7 is an explanatory diagram showing how the selected related information RT is added to the question information QT of the artificial intelligence system 100 according to an embodiment of the present invention and transmitted to the pre-trained artificial intelligence means; FIG. 8 is an explanatory diagram showing how the question information QT, which is a prompt of the artificial intelligence system 100 according to an embodiment of the present invention, is converted into vector coordinate system question data VQT; FIG. 9 is an explanatory diagram showing how the related information RT acquired by the artificial intelligence system 100 according to an embodiment of the present invention is converted into vector coordinate system related data VRT; FIG. 10 is an explanatory diagram showing how the vector coordinate system related data VRT of the artificial intelligence system 100 according to an embodiment of the present invention is stored in the vector index 123; FIG. 11(A) is an explanatory diagram of the vector coordinate system question data VQT of the artificial intelligence system 100 according to an embodiment of the present invention; FIG. 11(B) is an explanatory diagram of the correlation search between the vector coordinate system question data VQT and the vector coordinate system related data VRT of the artificial intelligence system 100 according to an embodiment of the present invention;FIG. 13 is an explanatory diagram showing a first modified example of how selected related information RT is added to question information QT of the artificial intelligence system 100 according to an embodiment of the present invention and transmitted to the pre-trained artificial intelligence means; FIG. 14 is a diagram showing an example of a reference range specification window 151 of the browser of the administrator terminal 150 of the artificial intelligence system 100 according to an embodiment of the present invention;

[0023] As shown in FIG. 1, an artificial intelligence system 100 according to an embodiment of the present invention includes a user terminal 110, a first server 120 and a second server 130 which are examples of servers, and a third server 140 on the Internet. Of these, the first server 120 has a bot 121 which is also a display program. On the other hand, the second server 130 has a chat GPT, which is an example of a text generation means 131 of a pre-trained artificial intelligence means. As a technical concept, the first server 120, the second server 130, and the third server 140 may have a common physical configuration, or may have physically different configurations. The user terminal 110 is provided so as to be able to communicate freely with the chat GPT, which is an example of the text generation means 131 of the pre-trained artificial intelligence means of the second server 130, via the first server 120.

[0024] More specifically, the user terminal 110 communicates with the first server 120 , and the browser 111 of the user terminal 110 displays a talk screen 112 . The talk screen 112 has a talk display area 113 that displays question information QT and generated text data GT, which are exchanges between the user and the pre-trained artificial intelligence means. Then, suppose that text is input in the talk display area 113 of the talk screen 112 and a send operation is performed.

[0025] At this time, the user terminal 110 transmits the text data to the first server 120 as question information QT. The first server 120 then acquires related information RT related to the received question information QT from the Internet. Alternatively, when text is input and a send operation is performed, the browser 111 of the user terminal 110 obtains related information RT related to the question information QT from the Internet as the text data as question information QT, and sends the question information QT and related information RT to the first server 120.

[0026] Then, the first server 120 generates text data of the content of the question information QT based on the related information RT and the question information QT using chat GPT, which is an example of the text generation means 131, which is a pre-trained artificial intelligence means. Furthermore, the first server 120 transmits the generated text data GT obtained through the execution to the user terminal 110. Then, the user terminal 110 is configured to display the received generated text data GT in the talk display area 113 of the talk screen 112 of the browser 111.

[0027] As a result, related information RT is acquired from the Internet in real time and learned by the so-called RAG (Retrieval Augmented Generation), which is an external information source for the artificial intelligence system 100. Generated text data GT is then generated as an answer based on the question information QT and related information RT and displayed on the talk screen 112. As a result, users can get answers based on relevant information RT, which is the latest information such as current affairs news, in real time.

[0028] Next, an example of the operation of the artificial intelligence system 100 will be described in more detail. As shown in FIG. 2, in step S1, which is an access step, the user terminal 110 or the first server 120 determines whether or not the user terminal 110 has accessed the first server 120. More specifically, when the user terminal 110 accesses the first server 120 and makes an HTTP request, the first server 120 transmits data of the talk screen 112 (see FIG. 3) to the user terminal 110 as an HTTP response.

[0029] Therefore, whether or not there is access can be determined by either the user terminal 110 or the first server 120. If it is determined that there has been an access, the process proceeds to step S2, and if it is determined that there has not been an access, step S1 is repeated. The determination of whether or not there is access may be made based on whether or not the user has logged in to the artificial intelligence system 100.

[0030] In step S2, as shown in FIG. 3, as a talk screen display step, the browser 111 of the user terminal 110 displays the talk screen 112 based on the HTTP response. Here, the talk screen 112 has a talk display area 113 that displays question information QT and generated text data GT, which are interactions between the user and the chat GPT, which is an example of a text generation means 131 of a pre-trained artificial intelligence means.

[0031] In step S3, which is a step for determining whether or not an instruction or command has been input, the user terminal 110 determines whether or not text has been entered into the input field 114 of the talk display area 113 of the talk screen 112 as instruction or command information or question information QT, which is a so-called prompt, and a sending operation has been performed. For example, suppose that a text "What are the advantages of electric vehicles?" is entered as an example of question information QT, and the send button is operated. If it is determined that an input / transmission operation has been performed in this way, the process proceeds to step S4, and if it is determined that no input / transmission operation has been performed yet, step S3 is repeated.

[0032] Regarding the method of inputting data such as instruction and command information and question information QT on the talk screen 112 of the browser 111 of the user terminal 110, text data may be input using a keyboard, or the microphone icon provided on the talk screen 112 may be operated to activate the microphone of the user terminal 110, and the browser 111 may convert the voice data input by voice into text data and input it.

[0033] In step S4, as a related information acquisition step, the user terminal 110 transmits text data to the bot 121 of the first server 120 as question information QT. For example, text data such as "What are the advantages of electric vehicles compared to gasoline vehicles?" is sent. Then, the first server 120 acquires related information RT related to the received question information QT from the Internet.

[0034] Specifically, the first server 120 inputs the question information QT into a search engine and acquires hits as search results as related information RT. The related information RT to be acquired may be one or more, and for example, the top two pieces of information are acquired from among the hits in the search results. Here, the order of ranking is, for example, the order in which the search engine determines that the relevance score is high. At this time, as shown in FIG. 4, for example, the first server 120 displays the message "Getting the latest information from the Internet (Web browsing) *This may take some time" on the browser 111 of the user terminal 110. Then, as described above, the first server 120 searches the Internet for the question information QT and acquires related information RT related to the question information QT from the third server 140, for example.

[0035] Alternatively, as a related information acquisition step, the browser 111 of the user terminal 110 acquires related information RT related to the question information QT as text data using a pre-installed browser 111 extension program, triggered by an input / send operation. Specifically, the browser 111 of the user terminal 110 inputs the question information QT into a search engine, and obtains hits as search results as related information RT. The related information RT to be acquired may be one or more, and for example, the top two pieces of information are acquired from among the hits in the search results. Here, the order of ranking is, for example, the order in which the search engine determines that the relevance score is high. Then, the browser 111 of the user terminal 110 transmits the question information QT and the related information RT to the first server 120.

[0036] In step S5, as a text data generation step (text generation and transmission step), the first server 120 generates text data of the content of the question information QT using chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, based on the related information RT and the question information QT. In step S6, as a generated text transmission step (text generation and transmission step), first server 120 transmits generated text data GT obtained by the execution to user terminal 110. For example, generated text data GT is transmitted that purports to say, "Since everything is electronically controlled, it is well suited to automated driving. Just enter your destination by voice and then depart..."

[0037] In step S7, as shown in FIG. 5, as a generated text display step, the user terminal 110 displays the received generated text data GT in the talk display area 113 of the talk screen 112 of the browser 111. For example, the user terminal 110 displays generated text data GT in the talk display area 113 of the talk screen 112, which purports to say, "Since everything is electronically controlled, it is compatible with automated driving. Just enter your destination by voice and then depart..."

[0038] As a result, as described above, related information RT is acquired from the Internet in real time and learned by the so-called RAG (Retrieval Augmented Generation), which is an external information source for the artificial intelligence system 100. Generated text data GT is then generated as an answer based on the question information QT and related information RT and displayed on the talk screen 112. As a result, users can get answers based on relevant information RT, which is the latest information such as current affairs news, in real time.

[0039] In addition, when chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, generates generated text data GT, it may include URL information of related information RT in the generated text data GT as basis information for the answer content. As shown in FIG. 5, the text portion of the evidence information is linked to the URL information of the website, which is the related information RT. The browser 111 may be configured so that when a text portion of the basis information is operated on, the browser 111 displays a page of the website that is the related information RT. This allows users to simply click to display the area they want to check. As a result, users can easily check the part of the website page on the Internet that is the source of the answer they want to check.

[0040] Furthermore, in this embodiment, the talk screen 112 has a real-time internet reference button 115 as shown in FIGS. When a real-time internet reference button 115 is operated to switch reference ON, and then text is input and a transmission operation is performed, the first server 120 acquires related information RT related to the received question information QT from the internet.

[0041] Alternatively, after the reference is switched ON, when text is input and a send operation is performed, the browser 111 of the user terminal 110 acquires the text data as question information QT and related information RT related to the question information QT from the Internet, and transmits the question information QT and related information RT to the server. Then, the first server 120 generates text data of the content of the question information QT using chat GPT, which is an example of text generation means 131 that is a pre-trained artificial intelligence means, based on the related information RT and the question information QT.

[0042] On the other hand, when the real-time Internet reference button 115 is not operated and reference is OFF, and text is input and a send operation is performed, the first server 120 does not obtain related information RT related to the received question information QT from the Internet, but instead generates text data of the content of the question information QT using chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means of the second server 130.

[0043] Alternatively, when reference is OFF and text is entered and a send operation is performed, the browser 111 of the user terminal 110 sends the text data as question information QT to the first server 120 without obtaining related information RT related to the question information QT from the Internet. The first server 120 is configured to generate text data of the content of the question information QT using chat GPT, which is an example of the text generation means 131, which is the pre-trained artificial intelligence means of the second server 130.

[0044] This allows the user to freely select whether or not to acquire related information RT in real time. As a result, when reference is OFF, it takes less time to obtain related information RT and emphasis is placed on response, while when reference is ON, it takes more time to obtain related information RT, but the latest related information RT is referenced and emphasis is placed on the freshness of the information.

[0045] In addition, in this embodiment, when the first server 120 obtains related information RT related to the received question information QT from the Internet, or when the browser 111 of the user terminal 110 obtains related information RT related to the question information QT from the Internet using text data as question information QT, the first server 120 or the browser 111 of the user terminal 110 refers to the update date and time information RN for the related information RT.

[0046] Then, as shown in FIG. 6(A), the first server 120 or the browser 111 of the user terminal 110 acquires one or more pieces of related information RT from the top, giving priority to the most recent information based on update date and time information RN. Alternatively, as shown in FIG. 6(B), the browser 111 of the first server 120 or the user terminal 110 is configured to obtain one or more pieces of related information RT for a predetermined period of time in the past based on the current date and time regarding the update date and time information RN.

[0047] For example, as described above, the first server 120 or the browser 111 of the user terminal 110 inputs the question information QT into a search engine, and obtains hits as search results as related information RT. The related information RT to be acquired may be one or more, and for example, the top two pieces of information are acquired from among the hits in the search results. At this time, a filter is used for the update date and time information RN, and one or more pieces of related information RT are acquired from the top, with priority given to the newest updated date and time information, as shown in FIG. 6(A). Alternatively, at this time, a filter is used for the update date and time information RN to obtain one or more pieces of related information RT for a predetermined period of time in the past based on the current date and time, as shown in FIG. 6(B).

[0048] As a result, the update date and time information RN is referenced, and relatively old information is not acquired from the related information RT on the Internet, but relatively new information is acquired, and the content of the generated text data GT as an answer to the question information QT does not contain relatively old information, but contains relatively new information. As a result, users can get answers based on relevant information RT, which is more up-to-date than current news, in real time.

[0049] As shown in FIG. 7, as described above, the user terminal 110 transmits text data to the bot 121 of the first server 120 as question information QT. Furthermore, in this embodiment, as shown in Figure 8, the first server 120 may be configured to perform a so-called embedding process, in which the language data of the question information QT is converted into vector coordinate system question data VQT using, for example, an embedding model, which is an example of a vector coordinate conversion means 122 provided in the first server 120. The vector coordinate conversion means 122 is provided to convert input data into data in a vector coordinate system having a plurality of parameters. For example, it determines vector components based on each element of the data to be converted and its intensity.

[0050] Here, the "data in the vector coordinate system" is, for example, vector data consisting of 1536 dimensions (parameters) of variables. In other words, it is vector data specified by 1536 components. The number of components of a vector may be any number. Moreover, although the vector coordinate conversion means 122 is provided in the first server 120 as an example, it may be provided in another server. The vector coordinate conversion means 122 may be provided in any of the first server 120, the second server 130, and the user terminal 110.

[0051] In addition, the first server 120 also acquires from the Internet a plurality of pieces of related information RT that are related to the received question information QT. Then, as shown in FIG. 9, the first server 120 converts the acquired plurality of pieces of related information RT into vector coordinate system related data VRT using vector coordinate conversion means 122, which is pre-trained artificial intelligence means. Then, as shown in FIG. 10, the first server 120, for example, associates the vector coordinate system related data VRT with the related information RT before conversion, maps it to the vector index 123, and stores it.

[0052] Furthermore, as shown in Figures 7, 11(A), and 11(B), the first server 120 searches for the vector coordinate system related data VRT that has the closest relationship to the vector of the vector coordinate system question data VQT, and performs a so-called correlation search. Then, the first server 120 selects the most closely related vector coordinate system related data VRT or related information RT corresponding to the most closely related vector coordinate system related data VRT.

[0053] Furthermore, the first server 120 performs text data generation using chat GPT, which is an example of a text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, based on the selected related information RT and question information QT. Then, the first server 120 transmits the generated text data GT obtained through the execution to the user terminal 110.

[0054] As shown in Figure 7 above, the first server 120 is configured to perform the embedding process and the correlation search, but instead of this configuration, as variant example 1, the user terminal 110 may be configured to perform the embedding process and the correlation search, as shown in Figure 12. More specifically, as shown in FIG. 8, the browser 111 of the user terminal 110 converts text data into question information QT into vector coordinate system question data VQT using the vector coordinate conversion means (122). Here, the vector coordinate conversion means (122) may be provided in the user terminal 110 or in the server. At the same time, the browser 111 of the user terminal 110 acquires multiple pieces of related information RT related to the question information QT from the Internet.

[0055] Then, the browser 111 of the user terminal 110 converts the acquired multiple pieces of related information RT into vector coordinate system related data VRT using a vector coordinate conversion means (122), which is a pre-trained artificial intelligence means, as shown in Figure 9. Then, as shown in FIG. 10, the browser 111 of the user terminal 110 associates the vector coordinate system related data VRT with the related information RT before conversion, maps it to the vector index 123, and stores it.

[0056] Furthermore, the browser 111 of the user terminal 110 searches for the vector coordinate system related data VRT that has the closest relationship to the vector of the vector coordinate system question data VQT, as shown in Figures 11(A) and 11(B), and performs a so-called correlation search. Then, the browser 111 of the user terminal 110 selects the most closely related vector coordinate system related data VRT or related information RT corresponding to the most closely related vector coordinate system related data VRT.

[0057] Furthermore, the browser 111 of the user terminal 110 adds the selected related information RT to the question information QT and transmits it to the first server 120. Then, the first server 120 generates text data based on the selected related information RT and question information QT using chat GPT, which is an example of a text generation means 131, which is a pre-trained artificial intelligence means of the second server 130. The first server 120 may then be configured to transmit the generated text data GT obtained through the execution to the user terminal 110.

[0058] As a result, in the artificial intelligence system 100, only related information RT that has a content that is highly relevant to the question is added to the question information QT and sent to the chat GPT, which is an example of a text generation means 131. As a result, the quality of the answers can be improved. Furthermore, only related information RT with content that is highly relevant to a selected small portion of the question from external information sources is added to the question information QT and sent to chat GPT, which is an example of text generation means 131, and the amount of data sent to chat GPT, which is an example of text generation means 131, is relatively small. As a result, it is possible to avoid a long response time from sending a question to receiving a reply.

[0059] As shown in FIG. 13, as a second variant, multiple pieces of related information RT may be acquired in the browser 111 of the user terminal 110, and the first server 120 may perform an embedding process on the question information QT and the multiple pieces of related information RT to perform a correlation search. More specifically, the browser 111 of the user terminal 110 acquires multiple pieces of related information RT related to the question information QT from the Internet. Then, the browser 111 of the user terminal 110 transmits the question information QT and the plurality of pieces of related information RT to the first server 120.

[0060] Then, as shown in FIG. 8, the first server 120 converts the text data into question information QT into vector coordinate system question data VQT using the vector coordinate conversion means 122. At the same time, the first server 120 converts the plurality of pieces of related information RT into vector coordinate system related data VRT using vector coordinate conversion means 122, which is pre-trained artificial intelligence means, as shown in FIG. Then, as shown in FIG. 10, the first server 120 associates the vector coordinate system related data VRT with the related information RT before conversion, maps it to the vector index 123, and stores it.

[0061] Furthermore, the first server 120 searches for the vector coordinate system related data VRT that has the closest relationship to the vector of the vector coordinate system question data VQT, as shown in Figures 11(A) and 11(B), and performs a so-called correlation search. Then, the first server 120 selects the most closely related vector coordinate system related data VRT or related information RT corresponding to the most closely related vector coordinate system related data VRT.

[0062] Furthermore, the first server 120 performs text data generation using chat GPT, which is an example of a text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, based on the selected related information RT and question information QT. The first server 120 may then be configured to transmit the generated text data GT obtained through the execution to the user terminal 110.

[0063] In this embodiment, the browser of the administrator terminal 150, which can communicate with the first server 120, displays a reference range designation window 151, as shown in FIG. In the reference range specification window 151, the reference range on the Internet when acquiring related information RT can be freely set by specifying the URL information of the website. In other words, when the first server 120 obtains related information RT related to the received question information QT from the Internet, or when the browser 111 of the user terminal 110 obtains related information RT related to the question information QT from the Internet using text data as question information QT, the related information RT is configured to be obtained within a specified range based on the URL information of a website previously specified on the administrator terminal 150, which can communicate with the first server 120.

[0064] In this way, by having the administrator specify in advance the range of reliable electronic information sources, related information RT can be acquired on the Internet within the range of the pre-specified electronic information sources. As a result, unreliable electronic information sources can be excluded and RAG learning can be performed in real time within the range of reliable electronic information sources, allowing users to obtain answers to question information QT based on RAG learning from reliable electronic information sources.

[0065] In addition to real-time RAG learning, it is of course possible to configure the administrator terminal 150, which can communicate with the first server 120, to acquire related information RT within a specified range based on at least one of electronic file information, website URL information, and database information, which are pre-specified electronic information sources, as advance RAG learning. That is, language character information is extracted from the electronic information source based on at least one of electronic file information, website URL information, and database information, which are electronic information sources specified on the administrator terminal 150 that can communicate with the first server 120. Then, the language character file data, which is data of the extracted language character information, is divided into a plurality of divided language character file data, and so-called chunking processing is performed. Furthermore, the divided post-division language character file data may be converted into vector coordinate system post-division language data using a vector coordinate conversion means 122, which is a pre-trained artificial intelligence means of the second server 130, and the converted vector coordinate system post-division language data may be associated with the post-division language character file data before conversion and stored in a vector index 123. This allows the administrator to simply specify an electronic information source, and the language and character information of the electronic information source is extracted and a division process known as chunking is performed, and then a vector conversion process known as embedding is performed, and the relationship before and after the vector conversion is associated, and a vector index saving process known as mapping is performed. As a result, data for so-called RAG (Retrieval Augmented Generation), which is an external information source for the artificial intelligence system 100, can be easily generated.

[0066] The artificial intelligence system 100 thus obtained as an embodiment of the present invention includes a user terminal 110 and a first server 120 and a second server 130, which are examples of servers. The user terminal 110 communicates with the first server 120. The browser 111 of the user terminal 110 displays a talk screen 112. The talk screen 112 has a talk display area 113 for displaying question information QT and generated text data GT, which are an exchange between a user and a pre-trained artificial intelligence means. When text is input and a send operation is performed in the talk display area 113 of the talk screen 112, the user terminal 110 transmits the text data to the first server 120 as question information QT. The first server 120 obtains related information RT related to the received question information QT from the Internet. Alternatively, the browser 111 of the user terminal 110 The first server 120 acquires text data from the internet as question information QT and related information RT associated with the question information QT, and transmits the question information QT and related information RT to the server. The first server 120 generates text data for the content of the question information QT using chat GPT, which is an example of text generation means 131, a pre-trained artificial intelligence means of the second server 130, based on the related information RT and the question information QT, and transmits the generated text data GT obtained by this generation to the user terminal 110. The user terminal 110 displays the received generated text data GT in the talk display area 113 of the talk screen 112 of the browser 111. This configuration allows the RAG to learn in real time, and the user can obtain answers in real time based on the related information RT, which is the latest information such as current news.

[0067] Furthermore, the talk screen 112 has a real-time internet reference button 115, and when the real-time internet reference button 115 is operated to switch to reference ON and then text is input and a send operation is performed, the first server 120 acquires related information RT related to the received question information QT from the Internet, or the browser 111 of the user terminal 110 acquires related information RT related to the question information QT from the Internet as text data for the question information QT and transmits the question information QT and related information RT to the first server 120, and the first server 120 generates text data of the content of the question information QT using the chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, based on the related information RT and the question information QT. When no operation is performed and reference is OFF, and text is input and a send operation is performed, the first server 120 does not obtain the related information RT related to the received question information QT from the Internet, or the browser 111 of the user terminal 110 does not obtain the related information RT related to the question information QT from the Internet as text data for the question information QT, but instead transmits the question information QT to the first server 120, and the first server 120 generates text data for the content of the question information QT using chat GPT, which is an example of text generation means 131, which is the pre-trained artificial intelligence means of the second server 130.As a result, when reference is OFF, it does not take time to obtain the related information RT, and emphasis is placed on response.On the other hand, when reference is ON, it takes time to obtain the related information RT, but the latest related information RT is referenced, and emphasis is placed on the freshness of the information.

[0068] Furthermore, when the first server 120 obtains the related information RT related to the received question information QT from the Internet, or when the browser 111 of the user terminal 110 obtains the related information RT related to the question information QT from the Internet using text data as the question information QT, it refers to the update date and time information RN for the related information RT and obtains one or more pieces of information from the top, with priority given to the most recent update date and time information RN, or obtains one or more pieces of information whose update date and time information RN is from a predetermined period in the past based on the current date and time.This allows the user to obtain answers based on the related information RT, which is the most up-to-date information in real time, more reliably than current news, etc.

[0069] Furthermore, the first server 120 converts the received question information QT into vector coordinate system question data VQT using the vector coordinate conversion means 122, acquires a plurality of pieces of related information RT related to the received question information QT from the Internet, converts the acquired plurality of pieces of related information RT into vector coordinate system related data VRT using the vector coordinate conversion means 122, searches for vector coordinate system related data VRT that is most closely related to the vector of the vector coordinate system question data VQT, and selects related information RT corresponding to the most closely related vector coordinate system related data VRT or the most closely related vector coordinate system related data VRT; alternatively, the browser 111 of the user terminal 110 converts the question information QT into vector coordinate system question data VQT using the vector coordinate conversion means 122 as text data, acquires a plurality of pieces of related information RT related to the question information QT from the Internet, converts the acquired plurality of pieces of related information RT into vector coordinate system related data VRT using the vector coordinate conversion means 122, and searches for vector coordinate system related data VRT that is most closely related to the vector of the vector coordinate system question data VQT. the first server 120 adds the selected related information RT to the question information QT, or the browser 111 of the user terminal 110 acquires a plurality of pieces of related information RT related to the question information QT from the Internet, and transmits the question information QT and the plurality of pieces of related information RT to the first server 120; the first server 120 converts the question information QT into vector coordinate system question data VQT using the vector coordinate conversion means 122, and also converts the plurality of pieces of related information RT into vector coordinate system related data VRT using the vector coordinate conversion means 122; the first server 120 searches for the vector coordinate system related data VRT that is most closely related to the vector of the vector coordinate system question data VQT, and selects the related information RT corresponding to the most closely related vector coordinate system related data VRT, or the plurality of pieces of related information VRT that are most closely related to the vector of the vector coordinate system question data VQT; and the first server 120 performs the following processing based on the selected related information RT and the question information QT:By using chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means, to generate text data, and transmitting the generated text data GT obtained by the generation to the user terminal 110, the quality of the answer can be improved and the response time from sending a question to receiving an answer can be prevented from becoming long.

[0070] Furthermore, when the first server 120 obtains related information RT related to the received question information QT from the Internet, or when the browser 111 of the user terminal 110 obtains related information RT related to the question information QT from the Internet using text data as question information QT, the related information RT is obtained within a specified range based on the URL information of a website, which is a pre-specified electronic information source, on the administrator terminal 150, which can communicate with the first server 120.This makes it possible to eliminate unreliable electronic information sources and perform RAG learning in real time within the range of reliable electronic information sources, and the user can obtain an answer to the question information QT from reliable electronic information sources based on RAG learning.

[0071] Furthermore, the program of the artificial intelligence system 100, which is an embodiment of the present invention, includes a talk screen display step S2 in which the user terminal 110 communicates with the first server 120, and the browser 111 of the user terminal 110 displays a talk screen 112 having a talk display area 113 that displays question information QT and generated text data GT, which are interactions between the user and the pre-trained artificial intelligence means of the second server 130; and when text is input and a send operation is performed in the talk display area 113 of the talk screen 112, the user terminal 110 transmits the text data to the server as question information QT, and the first server 120 obtains related information RT related to the received question information QT from the Internet, or the browser 111 of the user terminal 110 obtains related information RT related to the question information QT from the Internet as the text data as question information QT, By having a computer execute the following steps: a related information acquisition step S4 in which the question information QT and related information RT are sent to the server; text generation and transmission steps S5 and S6 in which the first server 120 generates text data for the question information QT using chat GPT, which is an example of text generation means 131, which is a pre-trained artificial intelligence means of the second server 130, based on the related information RT and the question information QT, and sends the generated text data GT obtained by the execution to the user terminal 110; and a generated text display step S7 in which the user terminal 110 displays the received generated text data GT in the talk display area 113 of the talk screen 112 of the browser 111, RAG can be trained in real time, and the effects are enormous, such as allowing users to obtain answers in real time based on the related information RT, which is the latest information such as current news. [Explanation of symbols]

[0072] 100 ··· Artificial Intelligence Systems 110 User terminal 111 ··· Browser 112 ··· Chat screen 113 ··· Talk display area 114 ··· Input field 115 ··· Real-time Internet browsing button 120 ··· First Server 121 ··· Bot 122 ··· Vector coordinate transformation means (pre-trained artificial intelligence means, embedding model) 123 ··· Vector index (spatial map) 130 Second Server 131 ··· Text generation methods (pre-trained AI methods, large-scale language models (LLM), chat GPT) 140 ··· Third Server (on the Internet) 150 ··· Administrator terminal 151 Reference Range Specification Window RT ··· Related Information RN... Update date and time information VRT: Vector coordinate system related data QT ··· Question information VQT Vector coordinate system query data GT ··· Generated text data

Claims

1. An artificial intelligence system comprising a user terminal and a server, wherein the user terminal can freely communicate with pre-trained artificial intelligence means of the server, The user terminal communicates with the server, and the browser of the user terminal displays a chat screen; The talk screen has a talk display area for displaying question information and generated text data, which are interactions between a user and the pre-trained artificial intelligence means, When text is input and a send operation is performed in the talk display area of ​​the talk screen, the user terminal transmits the text data as question information to the server, and the server acquires related information related to the received question information from the Internet, or the browser of the user terminal acquires related information related to the question information from the Internet as the text data as question information, and transmits the question information and the related information to the server, the server generates text data of the content of the question information using a text generation means which is a pre-trained artificial intelligence means based on the related information and the question information, and transmits the generated text data obtained by the execution to the user terminal; An artificial intelligence system characterized in that the user terminal is configured to display the received generated text data in the talk display area of ​​the browser's talk screen.

2. the talk screen has a real-time internet browsing button; When the real-time internet reference button is operated to switch reference ON and then text is input and a transmission operation is performed, the server acquires related information relating to the received question information from the internet, or the browser of the user terminal acquires related information relating to the question information from the internet as text data, and transmits the question information and related information to the server, and the server generates text data of the content of the question information using text generation means, which is pre-trained artificial intelligence means, based on the related information and question information; On the other hand, when the real-time Internet reference button is not operated and reference is OFF, and text is input and a send operation is performed, the server sends the question information to the server without obtaining related information related to the received question information from the Internet, or the browser of the user terminal does not obtain related information related to the question information from the Internet as text data, and the server generates text data of the content of the question information using a text generation means which is a pre-trained artificial intelligence means, as described in claim 1.

3. The artificial intelligence system described in claim 1 or claim 2 is characterized in that when the server obtains related information related to the received question information from the Internet, or when the browser of the user terminal obtains related information related to the question information from the Internet using text data as question information, it refers to the update date and time information of the related information and obtains one or more pieces from the top, prioritizing those with more recent update date and time information, or obtains one or more pieces of information with update date and time information from a predetermined period in the past based on the current date and time.

4. the server converts the received question information into vector coordinate system question data using vector coordinate conversion means, acquires a plurality of pieces of related information related to the received question information from the Internet, converts the acquired plurality of pieces of related information into vector coordinate system related data using the vector coordinate conversion means, searches for vector coordinate system related data that is most closely related to the vector of the vector coordinate system question data, and selects related information corresponding to the most closely related vector coordinate system related data or the top several pieces of related vector coordinate system related data that are most closely related; Alternatively, the browser of the user terminal converts the text data into vector coordinate system question data using the vector coordinate conversion means, obtains a plurality of pieces of related information related to the question information from the Internet, converts the obtained plurality of pieces of related information into vector coordinate system related data using the vector coordinate conversion means, searches for vector coordinate system related data that is most closely related to the vector of the vector coordinate system question data, selects the most closely related vector coordinate system related data or related information corresponding to the most closely related vector coordinate system related data in the top several rows, adds the selected related information to the question information, and transmits the result to the server; Alternatively, the browser of the user terminal acquires a plurality of pieces of related information related to the question information from the Internet, and transmits the question information and the plurality of pieces of related information to the server, and the server converts the question information into vector coordinate system question data using the vector coordinate conversion means, and converts the plurality of pieces of related information into vector coordinate system related data using the vector coordinate conversion means, searches for vector coordinate system related data that is most closely related to the vector of the vector coordinate system question data, and selects related information corresponding to the most closely related vector coordinate system related data or the top plurality of most closely related vector coordinate system related data, The artificial intelligence system described in claim 3, characterized in that the server is configured to generate text data using a text generation means which is a pre-trained artificial intelligence means based on the selected related information and question information, and to transmit the generated text data obtained by the generation to the user terminal.

5. The artificial intelligence system described in claim 4, characterized in that when the server obtains related information related to the received question information from the Internet, or when the browser of the user terminal obtains related information related to the question information from the Internet using text data as question information, it obtains related information within a specified range based on URL information of a website that is a pre-specified electronic information source on an administrator terminal that can communicate with the server.

6. A program for an artificial intelligence system in which a user terminal can communicate with a pre-trained artificial intelligence means of a server, a talk screen display step in which the user terminal communicates with the server, and a browser of the user terminal displays a talk screen having a talk display area for displaying question information and generated text data, which are interactions between the user and the pre-trained artificial intelligence means; a related information acquisition step in which, when text is input and a send operation is performed in the talk display area of ​​the talk screen, the user terminal transmits the text data as question information to a server, and the server acquires related information related to the received question information from the Internet, or the browser of the user terminal acquires related information related to the question information from the Internet as the text data as question information, and transmits the question information and the related information to the server; a text generation and transmission step in which the server generates text data of the content of the question information using a text generation means which is a pre-trained artificial intelligence means based on the related information and the question information, and transmits the generated text data obtained by the execution to the user terminal; A program for an artificial intelligence system that causes a computer to execute a generated text display step in which the user terminal displays the received generated text data in the talk display area of ​​the browser's talk screen.

Images