Artificial intelligence system and program of artificial intelligence system

Abstract

To provide an artificial intelligence system configured to improve the accuracy of making a response that matches the intent of a user.SOLUTION: An artificial intelligence system 100 includes a user terminal 110, and servers 120, 130, 140. The server 120 splits language text data of preset content information into multiple pieces of data, converts them into vector coordinate system split language data using vector coordinate conversion means 131, and stores the converted data in a vector index 121 in association with the data before and after the conversion. In response to an operation on the user terminal, the server 120 converts received question information into vector coordinate system question data using the vector coordinate conversion means 131, performs correlation search with the vector index, selects split language text data corresponding to vector coordinate system split language data having a closest relationship, to be added to the question information, executes text generation using text generation means 141, and transmits generated text data GT to the user terminal so as to be displayed on a browser.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an artificial intelligence system that includes a user terminal and a server, in which the user terminal displays or outputs audio based on the server, and a program for the system. [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] 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 a text-format summary directly from summary basic information in some language text format using a large-scale language model, so there was a risk that the large-scale language model would misrecognize the content of the summary basic information, and the accuracy of understanding the content of the summary basic information was insufficient.

[0005] Therefore, the present invention is intended to solve 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 thereof that can improve the accuracy of providing answers that are in line with the user's intentions. [Means for solving the problem]

[0006] The invention according to claim 1 is an artificial intelligence system comprising a user terminal and a server, the user terminal displaying or outputting audio based on the server, the server dividing language text data for content information specified by a preset document file or URL information into a plurality of pieces, converting the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means, and storing the converted vector coordinate system divided language data in association with the divided language text data before conversion in a vector index, and when a user inputs text on a browser of the user terminal, the user terminal transmits the text data to the server as question information, and the server receives the text data and converts it into a vector index. The problem described above is solved by converting the collected question information into vector coordinate system question data using a vector coordinate conversion means, searching for vector coordinate system divided language data in a vector index that is most closely related to the vector of the vector coordinate system question data, selecting the most closely related vector coordinate system divided language data or divided language text data corresponding to the top multiple most closely related vector coordinate system divided language data, having the server add the selected divided language text data to the question information, and generating text data using a text generation means which is a pre-trained artificial intelligence means, and transmitting the generated text data obtained by the execution to a user terminal, and having the user terminal display the received generated text data on a browser or output it as audio.

[0007] The invention according to claim 2 further comprises, in addition to the configuration of the artificial intelligence system according to claim 1, a plurality of pieces of content information specified by the preset document files or URL information, each of the plurality of content information specified by the plurality of document files or the plurality of URL information being regarded as one block, the content information specified by the plurality of document files or the plurality of URL information being a plurality of blocks, the server creates one piece of vector coordinate system data for each block based on the vector coordinate system division language data within the block, the server receives question information from the user terminal and converts it into vector coordinate system question data using a vector coordinate conversion means, and then searches for vector coordinate system data in each block that is most closely related to the vector of the vector coordinate system question data, and searches for the most closely related vector coordinate system data or the top multiple most closely related vector coordinate system data. The above-mentioned problem is further solved by the configuration in which a block corresponding to the vector coordinate system data is selected, for each selected block, the vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data in the vector index is searched for, the closest related vector coordinate system divided language data or multiple top closely related vector coordinate system divided language data is picked up in parallel processing from the selected block, and divided language text data corresponding to the picked closest related vector coordinate system divided language data or multiple top closely related vector coordinate system divided language data is selected, the server adds the selected divided language text data to the question information, and performs text data generation using a text generation means which is a pre-trained artificial intelligence means, and transmits the generated text data obtained by the execution to the user terminal.

[0008] The invention according to claim 3 further comprises, in addition to the configuration of the artificial intelligence system according to claim 1, a plurality of pieces of content information specified by the preset document file or URL information, each piece of content information specified by the plurality of document files or the plurality of URL information being regarded as one block, the content information specified by the plurality of document files or the plurality of URL information being a plurality of blocks, the server receives question information from the user terminal and converts it into vector coordinate system question data using a vector coordinate conversion means, and then searches for vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data in a vector index for each block, and sorts the closest vector coordinate system divided language data in each block or a top several closely related vector coordinate system divided language data. The above-mentioned problem is further solved by the configuration in which the closest related vector coordinate system divided language data picked up in all blocks by column processing, or the closest related vector coordinate system divided language data from among the multiple higher level closely related vector coordinate system divided language data, or the closest related vector coordinate system divided language data from among the multiple higher level closely related vector coordinate system divided language data, and the divided language text data corresponding to the closest related vector coordinate system divided language data picked up or the multiple higher level closely related vector coordinate system divided language data is selected, and the server adds the selected divided language text data to the question information, and performs text data generation using a text generation means which is a pre-trained artificial intelligence means, and transmits the generated text data obtained by the execution to the user terminal.

[0009] The invention of claim 4 further solves the above-mentioned problem by being configured in such a way that, in addition to the configuration of the artificial intelligence system described in claim 2 or claim 3, when content information identified by the predetermined document file or URL information is updated, the server detects that the update has been made based on update history information, and divides the language text data for the content information into multiple pieces only in blocks corresponding to the content information identified by the updated document file or URL information, converts the divided post-division language text data into vector coordinate system post-division language data using a vector coordinate conversion means, and associates the converted vector coordinate system post-division language data with the post-division language text data before conversion and stores them in vector indexes respectively.

[0010] The invention of claim 5 further solves the above-mentioned problem by being configured in such a way that, in addition to the configuration of the artificial intelligence system described in any one of claims 1 to 3, the server divides language text data regarding content information identified by a predetermined document file or URL information into multiple pieces, converts each of the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means, and displays on the browser of the user terminal the progress of storing the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion.

[0011] The invention according to claim 6 is a program for an artificial intelligence system in which a user terminal displays or outputs voice based on a server, the program comprising: a data conversion and storage step in which the server divides language text data for content information specified by a preset document file or URL information into a plurality of pieces, converts the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means, and stores the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion; a question transmission step in which the user terminal transmits the text data to the server as question information when a text input operation is performed by a user on a browser of the user terminal; and a question transmission step in which the server converts the received question information into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means. The above-mentioned problem is solved by having a correlation search step of converting the vector coordinate system question data into vector coordinate system question data using a conversion means, searching for vector coordinate system divided language data in a vector index that is most closely related to the vector of the vector coordinate system question data, and selecting divided language text data corresponding to the most closely related vector coordinate system divided language data or the top multiple most closely related vector coordinate system divided language data; an answer text generation and transmission step in which the server adds the selected divided language text data to the question information, generates text data using a text generation means which is a pre-trained artificial intelligence means, and transmits the generated text data obtained by the execution to a user terminal; and a generated text display and audio output step in which the user terminal displays the received generated text data on a browser or outputs it as audio. Effect 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 display or output audio based on the server, but also provides the following unique effects.

[0013] According to the artificial intelligence system of the invention of claim 1, language text data regarding content information identified by a predetermined document file or URL information is divided into multiple pieces, each of which is converted into vector coordinate system divided language data and stored in a vector index, question information entered by a user is also converted into vector coordinate system question data, vector coordinate system divided language data with a similar relationship is searched for, and divided language text data corresponding to the vector coordinate system divided language data is added to the question information, and generated text data is generated by a text generation means, which is a pre-trained artificial intelligence means, and displayed as an answer on the browser of the user terminal, thereby improving the accuracy of providing an answer that is in line with the user's intention. In other words, the language text data is converted into vector coordinate system data that is easy for a computer to compare (understand), a correlation search is performed on the vector coordinate system data for each divided part of the original document and each of the question information, and the divided part of the original document corresponding to the vector coordinate system data with a high correlation is added to the question information, and the text generation means, which is a pre-trained artificial intelligence means, does not need to read less relevant parts of the language text data, but only reads highly relevant parts and the question information to generate text data, thereby improving the accuracy of the answer.

[0014] According to the artificial intelligence system of the invention of claim 2, in addition to the effects of the invention of claim 1, when the number of content information identified by the preset document files or URL information is large, for example, 800 files or more, a correlation search is not performed in parallel processing within each block, but a correlation search is performed on the outline of each block, and a correlation search is performed only within the block with the highest correlation, or within the top several blocks in the correlation result ranking, and the vector coordinate system divided language data with the closest relationship thereamong, or the divided language text data corresponding to the top several closely related vector coordinate system divided language data, is selected and added to the question information, and text data generation is performed using a text generation means which is a pre-trained artificial intelligence means.Therefore, the response time from input of question information such as a user's question to the display of the answer can be significantly shortened compared to a configuration which performs serial processing, and interaction with the user can be smoother. Furthermore, compared to the configuration of claim 3, which performs correlation searches in parallel within all blocks, the processing time can be significantly shortened, thereby shortening the response time from inputting question information, such as a user's question, to displaying the answer.

[0015] According to the artificial intelligence system of the invention of claim 3, in addition to the effects of the invention of claim 1, when the number of content information identified by the preset document files or URL information is large, for example, 400 files or more, a correlation search is performed for each block by parallel processing, and the results of each block are selected from all blocks as a whole, and the divided language text data corresponding to the most closely related vector coordinate system divided language data, or the top multiple most closely related vector coordinate system divided language data, is selected and added to the question information, and text data generation is performed using a text generation means which is a pre-trained artificial intelligence means.Therefore, the response time from input of question information such as a user's question to the display of the answer can be significantly shortened compared to a configuration which performs serial processing, and interaction with the user can be smoother.

[0016] According to the artificial intelligence system of the invention according to claim 4, in addition to the effects achieved by the invention according to claim 2 or claim 3, when the content information specified by the document file or URL information is updated, the vector index is recreated only in the block corresponding to the updated content information. Therefore, compared with the configuration in which the vector index is recreated for all blocks in serial processing, the time required to recreate the vector index related to the update can be significantly shortened, and the risk of disturbing smooth communication with the user can be significantly reduced.

[0017] According to the artificial intelligence system of the invention according to claim 5, in addition to the effects achieved by the invention according to any one of claims 1 to 3, since the progress status is displayed on the browser, the user can roughly understand the accuracy of the answers to questions and inquiries about the content information specified by the preset document file or URL information. And, if necessary, the user can wait until the conversion and storage are completed in the progress status for the input timing of questions and inquiries, or re-enter them.

[0018] According to the program of the artificial intelligence system of the invention according to claim 6, similar to the effect achieved by the invention according to claim 1, the language text data about the content information specified by the preset document file or URL information is divided into a plurality of parts, each of which is converted into language data after being divided into a vector coordinate system and stored in the vector index. The question information input by the user is also converted into vector coordinate system question data. The language data after being divided into a vector coordinate system with a close relationship to this is searched, and the divided language text data corresponding to the language data after being divided into a vector coordinate system is added to the question information. Based on this, generated text data is generated by the text generation means, which is a pre-trained artificial intelligence means, and is displayed as an answer on the browser of the user terminal. Therefore, the accuracy of the answer according to the user's intention can be improved. In other words, the language text data is converted into vector coordinate system data that is easy for a computer to compare (understand), a correlation search is performed on the vector coordinate system data for each divided part of the original document and each of the question information, and the divided part of the original document corresponding to the vector coordinate system data with a high correlation is added to the question information, and the text generation means, which is a pre-trained artificial intelligence means, does not need to read less relevant parts of the language text data, but only reads highly relevant parts and the question information to generate text data, thereby improving the accuracy of the answer. [Brief description of the drawings]

[0019] [Figure 1] FIG. 1 is a diagram showing the concept of an artificial intelligence system according to an embodiment of the present invention. [Diagram 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. [Diagram 3] A figure showing an example of an AI document screen as a talk screen of a browser on a user terminal of an artificial intelligence system according to an embodiment of the present invention. [Figure 4] FIG. 2 is a diagram showing an example of a document registration window of a browser on a user terminal of an artificial intelligence system according to an embodiment of the present invention. [Diagram 5] FIG. 2 is an explanatory diagram showing how the artificial intelligence system according to an embodiment of the present invention divides language text data and converts the divided language text data into vector coordinate system divided language data. [Figure 6] A figure showing an example of interaction between a user and a pre-trained artificial intelligence means on the AI ​​document screen of the artificial intelligence system according to an embodiment of the present invention. [Figure 7] FIG. 2 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 8] 1A and 1B are explanatory diagrams of correlation search between vector coordinate system question data and vector coordinate system divided language data in an artificial intelligence system according to an embodiment of the present invention. [Figure 9]FIG. 13 is an explanatory diagram showing how selected segmented language text data is added to question information, which is a prompt of the artificial intelligence system according to an embodiment of the present invention, and transmitted to a pre-trained artificial intelligence means. [Figure 10] FIG. 2 is an explanatory diagram showing how parallel processing is performed for each block of the artificial intelligence system according to an embodiment of the present invention. [Figure 11] FIG. 11 is an explanatory diagram showing how language text data is divided and the divided language text data is converted into vector coordinate system divided language data when a document of the artificial intelligence system according to an embodiment of the present invention is updated. [Figure 12] FIG. 13 is a diagram showing an example of a progressive gauge on an AI document screen of an artificial intelligence system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] The artificial intelligence system of the present invention includes a user terminal and a server, in which the server divides language text data for content information specified by a preset document file or URL information into a plurality of pieces, converts the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means, and stores the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion, and when a user inputs text on a browser of the user terminal, the user terminal transmits the text data to the server as question information, and the server converts the received question information into vector coordinate system question data using the vector coordinate conversion means, The vector index is searched for vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data, and divided language text data corresponding to the closest related vector coordinate system divided language data or the top multiple most closely related vector coordinate system divided language data is selected, the server adds the selected divided language text data to the question information, and generates text data using a text generation means which is a pre-trained artificial intelligence means, and transmits the generated text data obtained by the execution to the user terminal, and the user terminal displays the received generated text data on a browser or outputs it as audio, so long as the accuracy of the answer being in line with the user's intention can be improved, any specific implementation is acceptable. Furthermore, the program for the artificial intelligence system of the present invention includes a data conversion and storage step in which the server divides language text data for content information specified by a preset document file or URL information into a plurality of pieces, converts the divided divided language text data into vector coordinate system divided language data using vector coordinate conversion means which is a pre-trained artificial intelligence means, and stores the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion; a question transmission step in which the user terminal transmits the text data to the server as question information when a text input operation is performed by the user on the browser of the user terminal; and a question transmission step in which the server converts the received question information into vector coordinate system question data using the vector coordinate conversion means, and stores the converted vector coordinate system divided language data in a vector index in association with the pre-conversion divided language text data. The specific implementation of the system may be any one as long as it has a correlation search step of searching for vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data and selecting divided language text data that corresponds to the closest related vector coordinate system divided language data or the top multiple closest related vector coordinate system divided language data, an answer text generation / transmission step of the server adding the selected divided language text data to the question information, generating text data using a text generation means which is a pre-trained artificial intelligence means, and transmitting the generated text data obtained by the execution to the user terminal, and a generated text display / audio output step of the user terminal displaying the received generated text data on a browser or outputting it as audio, thereby improving the accuracy of the answer that is in line with the user's intention.

[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 freely connected to a server via a communication network including a wide area network, such as the Internet, a local network, or a telephone line. Also, the servers may each be a single server or multiple servers on a cloud. The pre-trained artificial intelligence means may be composed of an embedding model, which is an example of a vector coordinate conversion means that converts input data such as language text data into data in a vector coordinate system, or an interactive text generation means that generates text data and is also called 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. In addition, a document file refers to an electronic data file whose contents include character information, and the data file format may be a simple text data format, a document data format, a cell-type spreadsheet file format, a pixel-type image file format such as a photo or video, etc., which allows characters to be read using an image recognition means or which allows characters to be read using a voice recognition means, or any format that allows characters to be directly recognized or indirectly recognized by converting voice to characters. The characters also include the source code of a program. Furthermore, the URL information (Uniform Resource Locator information), which is the location information for the specified information, may be any information that can identify a website page in HTML format or a document file as the specified information obtained by accessing a web server. EXAMPLES

[0022] An artificial intelligence system 100 according to an embodiment of the present invention will be described below with reference to FIGS. Here, FIG. 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 diagram 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 an AI document screen 112 as a talk screen 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(A) is a diagram showing an example of designating a document file DF in a document registration window 116 of the browser 111 of the user terminal 110 of the artificial intelligence system 100 according to an embodiment of the present invention, FIG. 4(B) is a diagram showing an example of designating predetermined information by URL information in a document registration window 116 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 an explanatory diagram showing a state in which language text data TX of the artificial intelligence system 100 according to an embodiment of the present invention is divided and the divided language text data DTX is converted into vector coordinate system divided language data VTX, and FIG. FIG. 7 is an explanatory diagram showing an example of an interaction between a user of the AI ​​document screen 112 of the intelligence system 100 and a chat GPT, which is an example of a text generation means 141 that is a pre-trained artificial intelligence means. FIG. 7 is an explanatory diagram showing how question information PT, 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 VPT. FIG. 8(A) is a diagram showing the vector coordinate system question data VPT of the artificial intelligence system 100 according to an embodiment of the present invention. FIG. 8(B) is an explanatory diagram showing a correlation search between the vector coordinate system question data VPT of FIG. 8(A) and the vector coordinate system divided language data VTX. FIG. 9 is an explanatory diagram showing how the selected divided language text data DTX is added to the question information PT, which is a prompt of the artificial intelligence system 100 according to an embodiment of the present invention, and transmitted to the chat GPT, which is an example of a text generation means 141 that is a pre-trained artificial intelligence means. FIG. 10 is an explanatory diagram showing how parallel processing is performed for each block BK of the artificial intelligence system 100 according to an embodiment of the present invention. FIG.FIG. 12 is an explanatory diagram showing how the language text data TX is divided and the divided language text data DTX is converted into vector coordinate system divided language data VTX when the document of the artificial intelligence system 100 according to the embodiment of the present invention is updated. FIG. 12 is a diagram showing an example of a progressive gauge 113b on the AI ​​document screen 112 of the artificial intelligence system 100 according to the 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, and a first server 120, a second server 130, and a third server 140, which are examples of servers. Of these, the first server 120 has a database, a vector index 121 which also serves as a spatial map, and a bot 122 which is also a display program, and the database contains information on the web page of the AI ​​document screen 112, which is the talk screen of the artificial intelligence system 100 displayed by the browser 111 of the user terminal 110. Here, the vector index 121 refers to a spatial map in a storage unit as a list, database, or concept for quickly finding a specific target from among a large number of vector coordinate system data.

[0024] The second server 130 also has vector coordinate conversion means 131 as pre-trained artificial intelligence means. The vector coordinate conversion means 131 is also called an embedding model, and is provided to convert input data into data in a vector coordinate system. Furthermore, the third server 140 has chat GPT, which is an example of a text generation means 141 also called a large-scale language model as 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.

[0025] Furthermore, the first server 120 divides (also called chunks) the language text data TX about the content information specified by a preset document file DF or URL information into a plurality of pieces. Then, the first server 120 converts the divided language text data DTX (also called chunk data) into vector coordinate system divided language data VTX using a vector coordinate conversion means 131, which is a pre-trained artificial intelligence means of the second server 130. At the same time, the first server 120 stores the converted vector coordinate system divided language data VTX in the vector index 121 in association with the pre-conversion divided language text data DTX. When the user inputs text into the browser 111 of the user terminal 110, the user terminal 110 transmits the text data to the first server 120 as question information PT.

[0026] Then, the first server 120 converts the received question information PT into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130. At the same time, the first server 120 searches the vector index 121 for the vector coordinate system divided language data VTX that is most closely related to the vector of the vector coordinate system question data VPT. Then, the first server 120 selects the most closely related vector coordinate system divided language data VTX, or the divided language text data DTX corresponding to the most closely related vector coordinate system divided language data VTX.

[0027] Next, the first server 120 adds the selected divided language text data DTX to the question information PT and performs text data generation using chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means of the third server 140. Then, 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 on the browser 111 or output it as voice.

[0028] As a result, the language text data TX regarding the content information identified by a preset document file DF or URL information is divided into multiple pieces, each of which is converted into vector coordinate system divided language data VTX and stored in the vector index 121, the question information PT entered by the user is also converted into vector coordinate system question data VPT, the vector coordinate system divided language data VTX that is closely related to this is searched for, and the divided language text data DTX corresponding to the vector coordinate system divided language data VTX is added to the question information PT, and based on this, generated text data GT is generated by chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means, and is displayed as an answer in the browser 111 of the user terminal 110. As a result, it is possible to improve the accuracy of providing answers that are in line with the user's intentions.

[0029] In other words, the language text data TX is converted into vector coordinate system data that is easy for a computer to compare (understand), a correlation search is performed on the vector coordinate system data (VTX, VPT) for each of the divided parts (DTX) of the original document and the question information PT, and the divided part (DTX) of the original document corresponding to the vector coordinate system data (VTX) with a high correlation is added to the question information PT, and the chat GPT, which is an example of a text generation means 141 that is a pre-trained artificial intelligence means, does not need to read less relevant parts in the language text data TX, but reads only the highly relevant parts and the question information PT to generate generated text data GT. As a result, the accuracy of the answers can be improved.

[0030] Next, an example of the operation of the artificial intelligence system 100 will be described in detail. As shown in FIG. 2, in step S1, as a document reference range setting determination step, first server 120 determines whether or not a document reference range has been set in the document file DF or the content information specified by the URL information. To set the document reference range, for example, first, the browser 111 is started on the user terminal 110 or the administrator terminal, and accesses the first server 120 based on the URL information of the artificial intelligence system 100 . Then, the user performs a login operation to the artificial intelligence system 100 using the first server 120 (administrator terminal).

[0031] Then, as shown in Figure 3, the browser 111 of the user terminal 110 displays an AI document screen 112, which is a talk screen that can be freely interacted with the chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means of the third server 140, via the first server 120. As an example, the AI ​​document screen 112 has a document item display area 113, a document content display area 114, and a talk display area 115. Of these, the document item display area 113 is provided to display a document item 113a. The document content display area 114 is provided to display the contents of the document file DF corresponding to the document item 113a selected in the document item display area 113 or the contents of the predetermined information specified by the URL information.

[0032] Furthermore, the talk display area 115 is configured to display interactions between the user and chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means of the third server 140, via the bot 122 of the first server 120. The talk display area 115 is configured like a chat screen. Here, the chat talk screen refers to a screen that displays exchanges of text and file data with one or more partners (including the bot 122) in chronological order on a single screen.

[0033] Next, assume that, on the AI ​​document screen 112 displayed by the browser 111, the user operates the "+New Document" section in the document item display area 113, for example. Then, as shown in FIG. 4(A), the browser 111 displays a pop-up document registration window 116. The document registration window 116 is provided with a "File Upload" tab and a "URL" tab. When the "File Upload" tab is selected, the document registration window 116 displays a drag area 116a as a predetermined area.

[0034] Then, for example, the user uses a mouse to drag the icon of a document file DF, for example, “The Appeal of Electric Vehicles”.pptx, from the drive of the memory section of the user terminal 110 or the file server to the drag area 116a, and operates the send button. In this way, when the icon of the document file DF is dragged into the drag area 116a as the specified area and the send button is operated, it is determined that the document reference range has been set and the process proceeds to step S2; on the other hand, if it is determined that the range has not yet been set, step S1 is repeated.

[0035] When the drag area 116a is clicked, the browser 111 pops up a file selection window, and a list of files is displayed in the file selection window. A desired document file DF may be selected and specified from this file list. Furthermore, as shown in FIG. 4(B), when the "URL" tab is selected in the document registration window 116, the document registration window 116 displays a URL input field 116b and a document name input field. The URL information, which is the location information for the predetermined information, may be specified by inputting the URL information in this URL input field 116b and operating the send button, or the file name may be input in the document name input field and the send button may be operated.

[0036] As for how to set the document reference range, for example, the administrator may drag the icons of multiple document files DF in the document registration window 116 displayed on the browser 111 of the administrator terminal to set multiple document files DF as the document reference range, or the document reference range may be set by specifying a folder in which multiple document files DF are stored in a file server or in the database of the first server 120.

[0037] In step S2, as a data conversion and storage step, the first server 120 extracts language text data TX for content information identified by a preset document file DF or URL information from the document file DF or the website identified by the URL information. Then, as shown in FIG. 5, the first server 120 divides the extracted language text data TX into a plurality of data, for example, each of which has 500 characters. Each sentence may be divided into a plurality of parts.

[0038] Then, the first server 120 converts the divided post-division language text data DTX into vector coordinate system post-division language data VTX using the vector coordinate conversion means 131 which is the pre-trained artificial intelligence means of the second server 130. At the same time, the first server 120 stores the converted vector coordinate system divided language data VTX in the vector index 121 in association with the pre-conversion divided language text data DTX.

[0039] In step S3, as a question operation determination step (question transmission step), the user terminal 110 determines whether or not the user has performed a text input operation and a transmission operation on the AI ​​document screen 112 of the browser 111 of the user terminal 110. More specifically, as shown in FIG. 6, it is determined whether or not a text is entered in the input field 115a in the talk display area 115 and the send button is operated. Note that data input by voice through operation of microphone icon button 115b may be converted by browser 111 into text data and input into input field 115a.

[0040] Also, for example, chat GPT, which is an example of text generation means 141, which is a pre-trained artificial intelligence means of the third server 140, generates multiple question information in which information regarding content information identified by the document file DF or URL information, which is the reference range, becomes answer information. Next, the bot 122 of the first server 120 transmits the generated question information to the user terminal 110. The generated question information is then displayed in the talk display area 115 of the AI ​​document screen 112 of the browser 111.

[0041] Here, when the displayed generated question information is clicked on, an example of the content of the generated question information, such as "What are the advantages compared to gasoline-powered vehicles?", is entered into input field 115a, and is transmitted to first server 120 as question information PT. When this click operation occurs, user terminal 110 determines that a text input operation has occurred and that a send operation has occurred. In this way, if it is determined that a text input operation and a transmission operation have been performed, the process proceeds to step S4, whereas if it is determined that no operation has been performed yet, step S3 is repeated.

[0042] In step S4, as shown in FIG. 6, as a question transmission step, user terminal 110 transmits text data to first server 120 as question information PT, which is an example of instruction command information also called a prompt. In step S5, as shown in FIG. 7, as a query vector coordinate conversion step (correlation search step), the first server 120 converts the received query information PT into vector coordinate system query data VPT using the vector coordinate conversion means 131 of the second server 130. FIG. 8(A) shows question information PT in text format data converted into vector coordinate system question data VPT. This representation is, for example, two-dimensional, but the actual vector coordinates are, for example, 1536-dimensional.

[0043] In step S6, as shown in FIG. 8(B), as a correlation search step, the first server 120 searches for the vector coordinate system divided language data VTX in the vector index 121 that has the closest relationship to the vector of the vector coordinate system question data VPT (see FIG. 8(A)). Then, the divided language text data DTX corresponding to the closest related vector coordinate system divided language data VTX or the top multiple closest related vector coordinate system divided language data VTX is selected.

[0044] In step S7, as shown in FIG. 9, as an answer text generating and transmitting step, the first server 120 adds the selected divided language text data DTX to the question information PT. Then, the first server 120 transmits data in which the divided language text data DTX has been added to the question information PT to the chat GPT, which is an example of the text generating means 141 that is the pre-trained artificial intelligence means of the third server 140.

[0045] In step S8, as an answer text generating and transmitting step, the first server 120 executes text data generation using chat GPT, which is an example of the text generating means 141, which is the pre-trained artificial intelligence means of the third server 140. Then, the first server 120 transmits the generated text data GT obtained through the execution to the user terminal 110. In step S9, as shown in Figures 6 and 9, as a generated text display / audio output step, the user terminal 110 displays the received generated text data GT on the AI ​​document screen 112 of the browser 111, or outputs it as audio using a speaker.

[0046] As a result, as described above, the language text data TX regarding the content information identified by a preset document file DF or URL information is divided into multiple pieces, each of which is converted into vector coordinate system divided language data VTX and stored in the vector index 121, the question information PT entered by the user is also converted into vector coordinate system question data VPT, the vector coordinate system divided language data VTX that is closely related to this is searched for, and the divided language text data DTX corresponding to the vector coordinate system divided language data VTX is added to the question information PT, and based on this, generated text data GT is generated by the chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means, and is displayed as an answer in the browser 111 of the user terminal 110. As a result, it is possible to improve the accuracy of providing answers that are in line with the user's intentions.

[0047] In other words, the language text data TX is converted into vector coordinate system data that is easy for a computer to compare (understand), a correlation search is performed on the vector coordinate system data (VTX, VPT) for each of the divided parts (DTX) of the original document and the question information PT, and the divided part (DTX) of the original document corresponding to the vector coordinate system data (VTX) with a high correlation is added to the question information PT, and the chat GPT, which is an example of a text generation means 141 that is a pre-trained artificial intelligence means, does not need to read less relevant parts in the language text data TX, but reads only the highly relevant parts and the question information PT to generate generated text data GT. As a result, the accuracy of the answers can be improved.

[0048] In addition, at each exchange in the talk display area 115, a speaker icon button 115c is provided. When the speaker icon button 115c is operated, the contents of the exchange, such as the question information PT and the generated text data GT, are converted into voice data and output as voice using the speaker of the user terminal 110. This allows for viewing, speaking, and listening without the need for text input. As a result, the user can smoothly interact with chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means, while viewing the contents of the document file DF displayed in the document content display area 114 of the AI ​​document screen 112.

[0049] In addition, when chat GPT, which is an example of text generation means 141, which is a pre-trained artificial intelligence means, generates text data based on question information PT, chat GPT, which is an example of text generation means 141, which is a pre-trained artificial intelligence means, may be configured to generate basis information PF regarding which page or part of the content information identified by the loaded document file DF or URL information the text data is based on. Then, the bot 122 of the first server 120 transmits the generated text data GT and the basis information PF to the user terminal 110. As shown in FIG. 6, the user terminal 110 is configured to display the generated text data GT and the basis information PF in a talk display area 115 of an AI document screen 112 of a browser 111.

[0050] As a result, not only the generated text data GT but also the basis information PF are displayed in the talk display area 115 of the AI ​​document screen 112 of the browser 111. As a result, the user can easily check the accuracy and reliability (veracity) of the generated text data GT by referring to the content information specified by the original document file DF or URL information based on the evidence information PF. In other words, the user can easily confirm which part and what the text data was generated based on, without simply looking at the generated text data GT and accepting the contents of the generated text data GT.

[0051] In addition, when the basis information PF is clicked, the browser 111 is configured to highlight the portion of the document file DF displayed in the document content display area 114 that corresponds to the basis information PF, based on the link information set in the basis information PF. As a result, the portion of the document file DF that corresponds to the basis information PF is highlighted on the AI ​​document screen 112 of the browser 111. As a result, the user can easily check the accuracy and reliability (veracity) of the generated text data GT by referring to the highlighted portions of the original document file DF based on the evidence information PF.

[0052] Furthermore, in this embodiment, there are a plurality of pieces of content information specified by preset document files DF or URL information. As shown in FIG. 10, the content information specified by a plurality of document files DF or a plurality of pieces of URL information is defined as one block BK, and the content information specified by the plurality of document files DF or the plurality of pieces of URL information is present in a plurality of blocks BK. That is, a document reference range is set for each of the document items 113a in the document item display area 113. The content information of one set document reference range exists for multiple blocks BK.

[0053] Furthermore, the first server 120 receives the question information PT from the user terminal 110 and converts it into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130 . Thereafter, the first server 120 searches for the vector coordinate system divided language data VTX that is most closely related to the vector of the vector coordinate system question data VPT in the vector index 121 for each block BK. Then, the first server 120 picks up the closest related vector coordinate system divided language data VTX in each block BK, or a plurality of top-ranked closely related vector coordinate system divided language data VTX, by parallel processing.

[0054] Next, the first server 120 further picks up the closest related vector coordinate system divided language data VTX from the vector coordinate system divided language data VTX picked up in all blocks BK, or the closest related vector coordinate system divided language data VTX from among the multiple top closely related vector coordinate system divided language data VTX, or the closest related vector coordinate system divided language data VTX from among the multiple top closely related vector coordinate system divided language data VTX. Then, the first server 120 further selects the picked up most closely related vector coordinate system divided language data VTX, or the divided language text data DTX corresponding to the top multiple most closely related vector coordinate system divided language data VTX.

[0055] Next, the first server 120 adds the selected divided language text data DTX to the question information PT. Furthermore, the first server 120 performs text data generation using chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means of the third server 140, based on the data in which the divided language text data DTX is added to the question information PT. The first server 120 is configured to transmit the generated text data GT obtained by the execution to the user terminal 110.

[0056] As a result, when the number of content information identified by the preset document file DF or URL information is large, for example, 400 files or more, a correlation search is performed for each block BK in parallel processing, and the results of each block BK are looked at as a whole across all blocks BK, and the most closely related vector coordinate system divided language data VTX, or the divided language text data DTX corresponding to the top multiple most closely related vector coordinate system divided language data VTX, is selected and added to the question information PT, and text data generation is performed using chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means. As a result, the response time from input of question information PT such as a user's question to display of the answer can be significantly shorter than in a configuration in which serial processing is performed, allowing smoother communication with the user.

[0057] In addition, instead of a configuration in which a correlation search is performed in parallel processing within all blocks BK, a configuration may be used in which a correlation search is performed on an outline of each block BK, and a correlation search is performed only within the block BK with the highest correlation, or only within the top several blocks BK in the correlation result ranking. More specifically, the first server 120 creates one piece of vector coordinate system data for each block BK based on the vector coordinate system divided language data VTX in the block BK. For example, one vector coordinate system data is created using the average value of all the vector coordinate system divided language data VTX in the block BK. Then, the first server 120 receives the question information PT from the user terminal 110 and converts it into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130, and then searches for the vector coordinate system data in each block BK that has the closest relationship to the vector of the vector coordinate system question data VPT. Next, the block BK corresponding to the vector coordinate system data that is the closest to the block BK or the vector coordinate system data that is the closest to the block BK is selected.

[0058] Furthermore, for each selected block BK, the vector index 121 is searched for the vector coordinate system divided language data VTX that is most closely related to the vector of the vector coordinate system question data VPT. Next, the closest related vector coordinate system divided language data VTX in the selected block BK or a plurality of top closest related vector coordinate system divided language data VTX are picked up in parallel processing. Furthermore, the divided language text data DTX corresponding to the picked up closest related vector coordinate system divided language data VTX or the top multiple closest related vector coordinate system divided language data VTX is selected.

[0059] Next, the first server 120 adds the selected divided language text data DTX to the question information PT. Then, the first server 120 executes text data generation using chat GPT, which is an example of the text generation means 141 that is the pre-trained artificial intelligence means of the third server 140. Furthermore, the first server 120 may be configured to transmit the generated text data GT obtained by the execution to the user terminal 110.

[0060] As a result, when the number of content information identified by the preset document file DF or URL information is large, for example, 800 files or more, a correlation search is not performed in parallel within each block BK, but rather a correlation search is performed on the outline of each block BK, and a correlation search is performed only within the block BK with the highest correlation, or within the top multiple blocks BK in the correlation result ranking, and the closest related vector coordinate system divided language data VTX or the divided language text data DTX corresponding to the top multiple closely related vector coordinate system divided language data VTX is selected and added to the question information PT, and text data generation is performed using chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means. As a result, the response time from input of question information PT such as a user's question to display of the answer can be significantly shorter than in a configuration in which serial processing is performed, allowing smoother communication with the user. In other words, compared to a configuration in which a correlation search is performed in parallel within all blocks BK, the processing time can be significantly shortened, and the response time from input of question information PT such as a user's question to display of the answer can be shortened.

[0061] Also in this embodiment, as shown in FIG. 11, when content information identified by a preset document file DF or URL information is updated, the first server 120 detects that the information has been updated based on the update history information. Then, the first server 120 divides the language text data TX for the content information into a plurality of pieces only in the block BK corresponding to the content information specified by the updated document file DF or the URL information. Then, the first server 120 converts the divided language text data DTX into vector coordinate system divided language data VTX using the vector coordinate conversion means 131 of the second server 130. In addition, the first server 120 is configured to store the converted vector coordinate system divided language data VTX in the vector index 121 in association with the pre-conversion divided language text data DTX.

[0062] As a result, when the content information specified by the document file DF or the URL information is updated, the vector index 121 is recreated only in the block BK corresponding to the updated content information. As a result, the time required to recreate the vector index 121 related to updating can be significantly shortened, as compared to a configuration in which the vector index 121 is recreated in all blocks BK by serial processing. This significantly reduces the risk of impeding smooth interaction with users.

[0063] Furthermore, in this embodiment, as described above, the first server 120 divides the language text data TX about the content information specified by the preset document file DF or URL information into a plurality of pieces. Then, the first server 120 converts the divided language text data DTX into vector coordinate system divided language data VTX using the vector coordinate conversion means 131 of the second server 130. At the same time, the first server 120 stores the converted vector coordinate system divided language data VTX in the vector index 121 in association with the pre-conversion divided language text data DTX. As shown in FIG. 12, the progress is displayed on a progressive gauge 113b on an AI document screen 112 of a browser 111 of a user terminal 110.

[0064] This allows the browser 111 to display the progress of the vector coordinate system data conversion, which is the so-called embedding. As a result, the user can roughly understand the accuracy of the answer to the question or inquiry about the content information specified by the preset document file DF or URL information. In addition, if necessary, you can wait until the conversion and saving is complete or re-enter your question or inquiry.

[0065] The artificial intelligence system 100, which is an embodiment of the present invention obtained in this manner, comprises a user terminal 110 and a first server 120, a second server 130 and a third server 140, which are examples of servers. The first server 120 divides language text data TX regarding content information specified by a preset document file DF or URL information into multiple pieces, and converts the divided divided language text data DTX into vector coordinate system divided language data VTX using a vector coordinate conversion means 131, which is a pre-trained artificial intelligence means of the second server 130. The converted vector coordinate system divided language data VTX is associated with the divided language text data DTX before conversion and stored in a vector index 121. When a user inputs text on the browser 111 of the user terminal 110, the user terminal 110 transmits the text data to the first server 120 as question information PT, and the first server 120 converts the received question information PT into a second The first server 120 converts the received data into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130, searches for vector coordinate system divided language data VTX that is closest to the vector of the vector coordinate system question data VPT in the vector index 121, and selects the closest related vector coordinate system divided language data VTX or divided language text data DTX corresponding to the top multiple closely related vector coordinate system divided language data VTX. The first server 120 adds the selected divided language text data DTX to the question information PT, and executes text data generation using chat GPT, which is an example of text generation means 141 that is a pre-trained artificial intelligence means of the third server 140, and transmits the generated text data GT obtained by the execution to the user terminal 110. The user terminal 110 displays the received generated text data GT on the browser 111 or outputs it as voice, thereby improving the accuracy of providing an answer that is in line with the user's intention.

[0066] In addition, there are a plurality of pieces of content information specified by a preset document file DF or URL information, and the content information specified by the plurality of document files DF or the plurality of URL information is regarded as one block BK, and there are a plurality of blocks BK of content information specified by the plurality of document files DF or the plurality of URL information, and the first server 120 creates one piece of vector coordinate system data for each block BK based on the vector coordinate system divided language data VTX in the block BK, and the first server 120 receives question information PT from the user terminal 110 and converts it into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130, and then searches for vector coordinate system data in each block BK that is closest to the vector of the vector coordinate system question data VPT, selects a block BK corresponding to the closest vector coordinate system data or multiple top most closely related vector coordinate system data, and selects the block BK that corresponds to the closest vector coordinate system data or the top most closely related vector coordinate system data for each selected block BK in the vector index 121. the first server 120 searches for the closest related vector coordinate system divided language data VTX, picks up the closest related vector coordinate system divided language data VTX or multiple top closely related vector coordinate system divided language data VTX in the selected block BK by parallel processing, selects divided language text data DTX corresponding to the picked up closest related vector coordinate system divided language data VTX or multiple top closely related vector coordinate system divided language data VTX, the first server 120 adds the selected divided language text data DTX to the question information PT, executes text data generation using chat GPT, which is an example of text generation means 141, which is the pre-trained artificial intelligence means of the third server 140, and transmits the generated text data GT obtained by execution to the user terminal 110. This configuration makes it possible to significantly shorten the response time from input of question information PT such as a question by a user to display of an answer, compared to a configuration in which serial processing is performed, and allows for smooth interaction with the user, compared to a configuration in which a correlation search is performed by parallel processing within all blocks BK,The processing time can be significantly shortened, and the response time from input of question information PT such as a user's question to display of the answer can be shortened.

[0067] Furthermore, there are a plurality of pieces of content information specified by a preset document file DF or URL information, and the content information specified by the plurality of document files DF or the plurality of URL information is regarded as one block BK, and there are a plurality of blocks BK of content information specified by the plurality of document files DF or the plurality of URL information, and the first server 120 receives question information PT from the user terminal 110 and converts it into vector coordinate system question data VPT using a vector coordinate conversion means 131 of the second server 130, and then searches for vector coordinate system divided language data VTX that is most closely related to the vector of the vector coordinate system question data VPT in the vector index 121 for each block BK, and picks up the most closely related vector coordinate system divided language data VTX or multiple top closely related vector coordinate system divided language data VTX in parallel processing, and the most closely related vector coordinate system divided language data VTX picked up in all blocks BK, Alternatively, the closest related vector coordinate system divided language data VTX or the closest related vector coordinate system divided language data VTX from among the multiple higher-level closely related vector coordinate system divided language data VTX is further picked up, and divided language text data DTX corresponding to the closest related vector coordinate system divided language data VTX or the multiple higher-level closely related vector coordinate system divided language data VTX is selected, and the first server 120 adds the selected divided language text data DTX to the question information PT, and performs text data generation using chat GPT, which is an example of a text generation means 141, which is a pre-trained artificial intelligence means of the third server 140, and transmits the generated text data GT obtained by the execution to the user terminal 110.By this configuration, the response time from input of question information PT such as a question by the user to the display of the answer can be significantly shortened compared to a configuration in which serial processing is performed, and interaction with the user can be smoother.

[0068] In addition, when content information specified by a preset document file DF or URL information is updated, the first server 120 detects that the update has been made based on the update history information, and divides the language text data TX for the content information into multiple pieces only in the block BK corresponding to the content information specified by the updated document file DF or URL information, and converts the divided divided language text data DTX into vector coordinate system divided language data VTX using the vector coordinate conversion means 131 of the second server 130, and associates the converted vector coordinate system divided language data VTX with the divided language text data DTX before conversion and stores them in the vector index 121, respectively.This configuration makes it possible to significantly shorten the time required to recreate the vector index 121 related to the update, compared to a configuration in which the vector index 121 is recreated in serial processing for all blocks BK, and significantly reduces the risk of hindering smooth communication with the user.

[0069] Furthermore, the first server 120 divides the language text data TX about the content information specified by a preset document file DF or URL information into multiple pieces, converts the divided divided language text data DTX into vector coordinate system divided language data VTX using the vector coordinate conversion means 131 of the second server 130, and associates the converted vector coordinate system divided language data VTX with the divided language text data DTX before conversion and stores them in the vector index 121, respectively.This configuration displays the progress on the progressive gauge 113b of the AI ​​document screen 112 of the browser 111 of the user terminal 110, allowing the user to roughly understand the accuracy of answers to questions and inquiries about the content information specified by the preset document file DF or URL information, and, if necessary, allows the user to wait until the conversion and saving are completed in the progress status, or to input the question or inquiry again.

[0070] Furthermore, the program of the artificial intelligence system 100 according to the embodiment of the present invention includes a data conversion and storage step S2 in which the first server 120 divides language text data TX on content information specified by a preset document file DF or URL information into a plurality of pieces, converts the divided divided language text data DTX into vector coordinate system divided language data VTX using a vector coordinate conversion means 131 which is a pre-trained artificial intelligence means of the second server 130, and stores the converted vector coordinate system divided language data VTX in the vector index 121 in association with the divided language text data DTX before conversion; question transmission steps S3 and S4 in which, when a text input operation is performed by the user on the browser 111 of the user terminal 110, the user terminal 110 transmits the text data to the first server 120 as question information PT; and the first server 120 converts the received question information PT into vector coordinate system question data VPT using the vector coordinate conversion means 131 of the second server 130. In addition, the system includes correlation search steps S5 and S6 in which the vector index 121 is searched for vector coordinate system divided language data VTX that is closest to the vector of the vector coordinate system question data VPT, and the system selects the closest related vector coordinate system divided language data VTX or divided language text data DTX corresponding to the higher level of closely related vector coordinate system divided language data VTX, answer text generation and transmission steps S7 and S8 in which the first server 120 adds the selected divided language text data DTX to the question information PT, and executes text data generation using chat GPT, which is an example of the text generation means 141 that is the pre-trained artificial intelligence means of the third server 140, and transmits the generated text data GT obtained by the execution to the user terminal 110, and a generated text display and voice output step S9 in which the user terminal 110 displays the received generated text data GT on the browser 111 or outputs it as voice, thereby providing a significant effect, such as improving the accuracy of providing an answer that is in line with the user's intention. [Explanation of symbols]

[0071] 100 ··· Artificial Intelligence Systems 110 User terminal 111 ··· Browser 112 ··· AI document screen (talk screen) 113 Document item display area 113a... Document item 113b Progressive gauge (progress of vector coordinate system data conversion) 114 Document content display area 115 ··· Talk display area 115a··· Input field 115b Microphone icon button 115c··· Speaker icon button 116 Document Registration Window 116a... Drag area 116b... URL input field 120 First Server (Server) 121 ··· Vector Index (Spatial Map) 122 Bot 130 Second Server (Server) 131 ··· Vector coordinate transformation means (pre-trained artificial intelligence means, embedding model) 140 Third Server (Server) 141 ··· Text generation means (pre-trained AI means, Chat GPT) DF Document file TX: (referent) language text data DTX: Segmented language text data VTX: Vector coordinate system divided language data PT ··· Question information VPT: Vector coordinate system query data GT: Generated text data PF: Basis information BK: Block

Claims

1. An artificial intelligence system comprising a user terminal and a server, the user terminal displaying or outputting audio based on the server, The server divides language text data for content information specified by a preset document file or URL information into a plurality of pieces, converts the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means, and stores the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion, When a user inputs text into a browser of the user terminal, the user terminal transmits the text data as question information to a server; the server converts the received question information into vector coordinate system question data using a vector coordinate conversion means, searches for vector coordinate system divided language data that is most closely related to the vector of the vector coordinate system question data in the vector index, and selects divided language text data corresponding to the most closely related vector coordinate system divided language data or the top multiple most closely related vector coordinate system divided language data; The server adds the selected divided language text data to the question information, and executes text data generation using a text generation means which is a pre-trained artificial intelligence means, 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 on a browser or output it as voice.

2. There are a plurality of pieces of content information specified by the preset document files or URL information, and the content information specified by the plurality of document files or the plurality of URL information is regarded as one block, and the content information specified by the plurality of document files or the plurality of URL information is a plurality of blocks, The server creates one vector coordinate system data for each block based on the vector coordinate system divided language data in the block, the server receives question information from the user terminal and converts it into vector coordinate system question data using a vector coordinate conversion means, searches for vector coordinate system data in each block that is closest to the vector of the vector coordinate system question data, and selects a block corresponding to the closest vector coordinate system data or a plurality of top most closely related vector coordinate system data; For each selected block, search for vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data in the vector index, and pick up the closest vector coordinate system divided language data or multiple top closest vector coordinate system divided language data in the selected block by parallel processing; Select the picked closest related vector coordinate system divided language data or divided language text data corresponding to the top multiple closest related vector coordinate system divided language data, The artificial intelligence system described in claim 1, characterized in that the server is configured to add the selected post-division language text data to question information, and generate text data using a text generation means which is a pre-trained artificial intelligence means, and transmit the generated text data obtained by the execution to a user terminal.

3. There are a plurality of pieces of content information specified by the preset document files or URL information, and the content information specified by the plurality of document files or the plurality of URL information is regarded as one block, and the content information specified by the plurality of document files or the plurality of URL information is a plurality of blocks, the server receives question information from the user terminal and converts it into vector coordinate system question data using a vector coordinate conversion means, searches for vector coordinate system divided language data that is closest to the vector of the vector coordinate system question data in the vector index for each block, and picks up the closest vector coordinate system divided language data in each block or a plurality of top most closely related vector coordinate system divided language data by parallel processing; Further picking up the closest related vector coordinate system divided language data or the closest related vector coordinate system divided language data from among the most closely related vector coordinate system divided language data in all blocks, or the most closely related vector coordinate system divided language data, and selecting divided language text data corresponding to the most closely related vector coordinate system divided language data or the most closely related vector coordinate system divided language data; The artificial intelligence system described in claim 1, characterized in that the server is configured to add the selected post-division language text data to question information, and generate text data using a text generation means which is a pre-trained artificial intelligence means, and transmit the generated text data obtained by the execution to a user terminal.

4. The artificial intelligence system described in claim 2 or claim 3, characterized in that when content information identified by the predetermined document file or URL information is updated, the server detects that the update has been made based on update history information, and divides the language text data for the content information into multiple pieces only in blocks corresponding to the content information identified by the updated document file or URL information, converts the divided post-division language text data into vector coordinate system post-division language data using a vector coordinate conversion means, and associates the converted vector coordinate system post-division language data with the post-division language text data before conversion and stores them in vector indexes.

5. The artificial intelligence system described in any one of claims 1 to 3, characterized in that the server is configured to divide language text data for content information identified by a predetermined document file or URL information into multiple pieces, convert each of the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means, and display on the browser of the user terminal the progress of storing the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion.

6. A program of an artificial intelligence system that a user terminal displays or outputs voice based on a server, a data conversion and storage step in which the server divides language text data for content information specified by a preset document file or URL information into a plurality of parts, converts the divided divided language text data into vector coordinate system divided language data using a vector coordinate conversion means which is a pre-trained artificial intelligence means, and stores the converted vector coordinate system divided language data in a vector index in association with the divided language text data before conversion; a question transmission step of, when a user inputs text on a browser of the user terminal, transmitting text data as question information to a server by the user terminal; a correlation search step in which the server converts the received question information into vector coordinate system question data using a vector coordinate conversion means, searches for vector coordinate system divided language data that is most closely related to the vector of the vector coordinate system question data in the vector index, and selects the most closely related vector coordinate system divided language data or divided language text data corresponding to multiple top-ranked vector coordinate system divided language data that are most closely related; an answer text generating and transmitting step in which the server adds the selected divided language text data to question information, generates text data using a text generating means which is a pre-trained artificial intelligence means, and transmits the generated text data to a user terminal; A program for an artificial intelligence system, comprising a generated text display / voice output step in which the user terminal displays the received generated text data on a browser or outputs it as voice.

Images