Information processing device, information processing method, and information processing system

Abstract

Provided is an information processing device using a dialog model which is easy for a user to use. The information processing device includes a reception unit, a processing unit, and an output unit. The reception unit receives first data from a computer via a first network. The processing unit is configured to execute: a process for acquiring, by using the first data, first text data corresponding to an instruction sentence, and second text data corresponding to at least another partial sentence; a process for acquiring, by inputting the first text data to a natural language processing model, third text data corresponding to a correction sentence of the instruction sentence; and a process for acquiring, by inputting the second text data and the third text data to the dialog model via a second network, fourth text data corresponding to a response sentence. The output unit outputs, to the computer via the first network, second data based on the fourth text data.

Description

Information processing device, information processing method, and information processing system

[0001] One aspect of the present invention relates to an information processing device, an information processing method, and an information processing system. Specifically, one aspect of the present invention relates to an information processing device, an information processing method, and an information processing system that can utilize processing using a dialogue model.

[0002] One embodiment of the present invention is not limited to the above technical field, and examples of the technical field of one embodiment of the present invention include semiconductor devices, display devices, light-emitting devices, power storage devices, memory devices, electronic devices, lighting devices, input devices (e.g., touch sensors), input / output devices (e.g., touch panels), driving methods thereof, and manufacturing methods thereof.

[0003] In recent years, the development of language models using neural networks has been actively carried out, and large-scale language models (LLMs) in particular have attracted attention. A large-scale language model is a natural language processing model trained using a large amount of data. A large-scale language model can realize, for example, a dialogue model that responds to user instructions. Non-Patent Document 1 discloses GPT-4 (Generative Pre-trained Transformer 4) as a large-scale language model and ChatGPT as a dialogue model. Patent Document 1 also discloses a dialogue model training method.

[0004] Special Publication No. 2023-508062

[0005] Summary of ChatGPT / GPT-4 Research and Perspective Towards the Future of Large Language Models, Yiheng Liu et al. (Submitted on 4 Apr 2023, [online], Internet <URL: https: / / arxiv.org / abs / 2304.01852>

[0006] When a dialogue model is given an input sentence called a prompt, it generates a response based on that input sentence. If an input sentence suitable for the dialogue model can be entered, it is easy to obtain the desired answer from the dialogue model. However, learning how to create the optimal input sentence that will obtain the desired answer requires repeated trial and error. Therefore, there is a demand for services that use dialogue models that are easy for users to use.

[0007] In addition, the sentences input to the dialogue model may be used for training. For example, if personal information or confidential information is input to the dialogue model, this information may be collected and used to train the dialogue model, which may lead to information leakage.

[0008] There is also concern about prompt injection, where a user inputs malicious prompts into a dialogue model, resulting in the model being misused. Specifically, the dialogue model may disclose information to the user that was not intended by the provider of the service using the dialogue model, or output inappropriate responses to the user.

[0009] In view of the above, an object of one embodiment of the present invention is to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is easy for users to use.Another object of one embodiment of the present invention is to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is safe for users to use.Another object of one embodiment of the present invention is to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is capable of preventing prompt injection.

[0010] Another object of one embodiment of the present invention is to provide a highly convenient information processing device, information processing method, or information processing system.Another object of one embodiment of the present invention is to provide a highly secure information processing device, information processing method, or information processing system.Another object of one embodiment of the present invention is to provide a novel information processing device, information processing method, or information processing system.

[0011] Note that the description of these problems does not preclude the existence of other problems. One embodiment of the present invention does not necessarily have to solve all of these problems. Problems other than these can be extracted from the description in the specification, drawings, and claims.

[0012] One aspect of the present invention is an information processing device having a reception unit, a processing unit, and an output unit, wherein the reception unit receives first data from a computer via a first network, the processing unit is configured to use the first data to obtain first text data corresponding to an instruction sentence and second text data corresponding to at least a portion of another sentence, a process of inputting the first text data into a natural language processing model to obtain third text data corresponding to a modified sentence of the instruction sentence, and a process of inputting the second text data and the third text data into a dialogue model via a second network to obtain fourth text data corresponding to a response sentence, and the output unit outputs second data based on the fourth text data to the computer via the first network.

[0013] Preferably, the first network is a local network and the second network is a global network, or the first network and the second network are each a global network.

[0014] Preferably, the processing unit inputs the first data into the judgment model to obtain a judgment result as to whether the first data can be input to the dialogue model, and further, if the judgment result is "no," acquires third data. Preferably, the judgment result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as criteria. The third data is data generated using the first data, for which the judgment result as to whether the first data can be input to the dialogue model is "yes," and the first text data and the second text data are acquired using the third data.

[0015] Preferably, the output unit outputs fourth data based on the third text data to the computer.

[0016] The third text data is preferably generated by modifying the first text data so that the third text data becomes an instruction sentence that is more appropriate for the dialogue model.

[0017] Another aspect of the present invention is an information processing method that receives first data from a computer via a first network, uses the first data to obtain first text data corresponding to an instruction sentence and second text data corresponding to at least a portion of another sentence, inputs the first text data into a natural language processing model to obtain third text data corresponding to a modified sentence of the instruction sentence, inputs the second text data and the third text data into a dialogue model via a second network to obtain fourth text data corresponding to a response sentence, and outputs second data based on the fourth text data to the computer via the first network.

[0018] Preferably, the first network is a local network and the second network is a global network. Alternatively, data is transmitted and received via a third network to and from a processing unit capable of executing processing using a natural language processing model, and the first and third networks are each local networks and the second network is a global network. Alternatively, data is transmitted and received via a third network to and from a processing unit capable of executing processing using a natural language processing model, and the first and second networks are each global networks and the third network is a local network.

[0019] It is preferable that a determination result as to whether the first data can be input to the dialogue model is obtained by inputting the first data to the determination model, and further, if the determination result is "no," third data is acquired. It is preferable that the determination result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as criteria. The third data is data generated using the first data, and the determination result as to whether the first data can be input to the dialogue model is "yes," and the first text data and the second text data are acquired using the third data.

[0020] It is preferable to output fourth data based on the third text data to the computer.

[0021] Another aspect of the present invention is an information processing system having a first computer, a second computer, and a third computer, wherein the second computer receives first data from the first computer via a first network, the second computer uses the first data to obtain first text data corresponding to an instruction sentence and second text data corresponding to at least a portion of another sentence, the second computer supplies the first text data to a natural language processing model to obtain third text data corresponding to a modified sentence of the instruction sentence, the third computer receives the second text data and the third text data from the second computer via the second network, the third computer uses the second text data and the third text data as input data and performs processing using a dialogue model to generate fourth text data corresponding to a response sentence, the third computer outputs the fourth text data to the second computer via the second network, and the second computer outputs second data based on the fourth text data to the first computer via the first network.

[0022] In the above information processing system, it is preferable that the first network is a local network and the second network is a global network, or that the first network and the second network are each a global network.

[0023] Preferably, the second computer inputs the first data into the judgment model to obtain a judgment result as to whether the first data can be input to the dialogue model, and further, if the judgment result is "no," acquires third data. Preferably, the judgment result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as criteria. The third data is data generated using the first data, and for which the judgment result as to whether the first data can be input to the dialogue model is "yes," and the first text data and the second text data are acquired using the third data.

[0024] The second computer preferably outputs fourth data based on the third text data to the first computer.

[0025] Another aspect of the present invention is an information processing system having a first computer and a second computer, wherein the first computer has a receiving unit, a first processing unit, and an output unit, and the second computer has a second processing unit, wherein the receiving unit is configured to receive first data from a third computer via a first network, and the first computer and the second computer transmit and receive data via a second network, wherein the second processing unit is configured to use the first data to execute a process of acquiring first text data corresponding to an instruction sentence and second text data corresponding to at least a portion of another sentence, and a process of acquiring third text data corresponding to a modified sentence of the instruction sentence by inputting the first text data into a natural language processing model, wherein the first processing unit is configured to execute a process of acquiring fourth text data corresponding to a response sentence by inputting the second text data and the third text data into a dialogue model via the third network, and the output unit outputs second data based on the fourth text data to the third computer via the first network.

[0026] Preferably, the first network and the second network are each a local network and the third network is a global network, or alternatively, the first network and the third network are each a global network and the second network is a local network.

[0027] According to one aspect of the present invention, it is possible to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is easy for users to use. Also, according to one aspect of the present invention, it is possible to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is safe for users to use. Also, according to one aspect of the present invention, it is possible to provide an information processing device, an information processing method, or an information processing system that uses a dialogue model and is capable of preventing prompt injection.

[0028] According to one embodiment of the present invention, a highly secure information processing device, an information processing method, or an information processing system can be provided. According to another embodiment of the present invention, a highly convenient information processing device, an information processing method, or an information processing system can be provided. According to another embodiment of the present invention, a novel information processing device, an information processing method, or an information processing system can be provided.

[0029] Note that the description of these effects does not preclude the existence of other effects. One embodiment of the present invention does not necessarily have all of these effects. Effects other than these can be extracted from the description in the specification, drawings, and claims.

[0030] FIG. 1 is a diagram showing an example of an information processing system. FIGS. 2A and 2B are diagrams showing an example of an information processing system. FIG. 3 is a block diagram showing an example of an information processing system. FIG. 4 is a diagram showing an example of an information processing system. FIGS. 5A and 5B are diagrams showing an example of an information processing system. FIG. 6 is a block diagram showing an example of an information processing system. FIG. 7 is a diagram showing an example of an information processing system. FIG. 8 is a diagram showing an example of an information processing system.

[0031] The embodiments will be described in detail with reference to the drawings. However, the present invention is not limited to the following description, and it will be readily understood by those skilled in the art that various changes can be made in form and detail without departing from the spirit and scope of the present invention. Therefore, the present invention should not be interpreted as being limited to the description of the embodiments shown below.

[0032] In the configuration of the invention described below, the same parts or parts having similar functions are denoted by the same reference numerals in different drawings, and repeated explanations thereof will be omitted. Furthermore, when referring to similar functions, the same hatching pattern may be used and no particular reference numeral may be assigned.

[0033] Furthermore, for ease of understanding, the position, size, range, etc. of each component shown in the drawings may not represent the actual position, size, range, etc. Therefore, the disclosed invention is not necessarily limited to the position, size, range, etc. disclosed in the drawings.

[0034] In this specification, the ordinal numbers "first" and "second" are used for convenience and do not limit the number of components or the order of the components (for example, the order of processes or the order of stacking). Furthermore, the ordinal numbers assigned to components in one part of this specification may not match the ordinal numbers assigned to the same components in other parts of this specification or in the claims.

[0035] Embodiment In this embodiment, an information processing system according to one embodiment of the present invention will be described with reference to FIGS.

[0036] An information processing system according to one embodiment of the present invention includes a first computer, a second computer, and a third computer. The first computer is an information terminal used by a user of the information processing system and can be referred to as a client computer. The third computer is capable of performing processing using a dialogue model that utilizes a large-scale language model. Because dialogue models that utilize a large-scale language model are extremely large, they are difficult to build within a local network and often must be used via a global network. Here, the first computer and the third computer are not directly connected via the global network, but are connected via a second computer. In other words, the second computer relays data transmission and reception between the first computer and the third computer.

[0037] The second computer does not input data received from the first computer directly to the third computer, but inputs data to the third computer after at least a portion of the data has been modified, converted, deleted, or the like. The data may be modified, converted, deleted, or the like by the second computer. In this case, the second computer may be capable of performing various processes using a natural language processing model. In this case, the second computer performs processes using a smaller model than the third computer. Therefore, the model used by the second computer can be built within a local network. Alternatively, the second computer may transmit data to another computer via the local network and receive data after at least one of modification, conversion, deletion, or the like.

[0038] The first computer and the second computer are capable of transmitting and receiving data via a first network, which may be a local network or a global network.

[0039] The second computer and the third computer can transmit and receive data via a second network, which is a global network.

[0040] The second computer receives first data from the first computer via the first network. The second computer then uses the first data to obtain first text data corresponding to the instruction sentence and second text data corresponding to at least a portion of another sentence. The second computer then supplies the first text data to a natural language processing model to obtain third text data corresponding to a correction of the instruction sentence.

[0041] The third computer receives the second text data and the third text data from the second computer via the second network. The third computer uses the second text data and the third text data as input data and performs processing using a dialogue model to generate fourth text data corresponding to a response sentence. The third computer then outputs the fourth text data to the second computer via the second network. Thereafter, the second computer outputs second data based on the fourth text data to the first computer via the first network.

[0042] Thus, in an information processing system according to one aspect of the present invention, first data is input from a first computer to a second computer, and second text data and third text data obtained based on the first data are input from the second computer to a third computer.

[0043] When first data is directly input from a first computer to a third computer, the user is required to consider and input an input sentence that is optimal for the dialogue model. Furthermore, if personal information, confidential information, or the like is input incorrectly, there is a risk of irreversible damage, which places a heavy burden on the user. On the other hand, in an information processing system according to one embodiment of the present invention, the data input by the user can be corrected using the second computer.

[0044] For example, the second computer can modify the instruction sentence entered by the user to make it more appropriate for the dialogue model, thereby obtaining a more appropriate answer from the dialogue model. In other words, even if the user does not know the optimal input sentence for the dialogue model, the desired answer can be more easily obtained from the dialogue model. This makes it possible to realize an information processing system that is easy for users to use.

[0045] Furthermore, for example, the second computer determines whether data entered by a user contains personal information, confidential information, malicious instructions, etc., and then deletes the personal information, confidential information, malicious instructions, etc. based on the determination result. This prevents personal information, confidential information, malicious instructions, etc. from being input into the dialogue model. Therefore, leakage of personal information, confidential information, etc. can be prevented, and an information processing system that can be used safely by users can be realized. Furthermore, for service providers using the information processing system, unintended information can be prevented from being disclosed to users, and a highly secure information processing system can be realized.

[0046] Furthermore, even if the first data is data other than text data, such as voice data or image data, the second computer can generate text data to be input to the third computer based on the first data, thereby improving the ease of use of the information processing system.

[0047] The second computer can also be regarded as an information processing device capable of executing processing using an information processing method according to one aspect of the present invention. Specifically, one aspect of the present invention is an information processing method that receives first data from a computer (corresponding to the first computer described above) via a first network, obtains first text data corresponding to an instruction sentence and second text data corresponding to at least a part of another sentence using the first data, inputs the first text data into a natural language processing model to obtain third text data corresponding to a correction sentence of the instruction sentence, inputs the second text data and the third text data into a dialogue model via a second network to obtain fourth text data corresponding to a response sentence, and outputs second data based on the fourth text data to the computer via the first network.

[0048] Furthermore, in the above information processing method, first data may be input to the judgment model to obtain a judgment result as to whether the first data can be input to the dialogue model, and if the judgment result is "no," third data may be acquired. The judgment result is, for example, a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as criteria. The third data is data generated using the first data, and for which the judgment result as to whether the first data can be input to the dialogue model is "yes." In this case, the above-mentioned first text data and second text data are acquired using the third data.

[0049] In the information processing method, fourth data based on the third text data may be output to the computer, allowing the user to check how the data he or she inputted has been corrected.

[0050] <Configuration Example 1 of Information Processing System> Fig. 1 shows an example of an information processing system according to this embodiment. In the information processing system shown in Fig. 1, an information processing device 10 is connected to an information processing device 40 via a network 30. The information processing device 10 is also connected to a plurality of information terminals 20 via a network 31. Fig. 1 shows information terminal 20a, information terminal 20b, information terminal 20c, and information terminal 20d as examples of the information terminals 20.

[0051] A user of the information processing system can access the information processing device 10 from the information terminals 20a to 20d, etc. The user can then receive a service using the information processing method according to one aspect of the present invention.

[0052] A user can use the information processing device 10, for example, by using dedicated software or an application installed on the information terminal 20. Alternatively, a user can use the information processing device 10, for example, from a web browser on the information terminal 20.

[0053] An organization such as a company that provides a service using an information processing method according to one embodiment of the present invention (hereinafter also referred to as a service provider) can provide the service using the information processing device 10. The service provider does not need to prepare the information processing device 40 on its own, but can use the information processing device 40 via the network 30.

[0054] The information processing device 10 is a device that can execute processing using an information processing method according to one aspect of the present invention. In FIG. 1, a server computer is illustrated as an example.

[0055] The information processing device 10 can perform information processing such as calculations using data input from the information terminal 20a via the network 31. The information processing device 10 can transmit the results of the information processing to the information terminal 20a via the network 31. This can reduce the calculation load on the information terminal 20a. Here, the information terminal 20a has been described as an example, but the same can be said for the information terminals 20b to 20d.

[0056] The information processing device 10 can input data to the information processing device 40 via the network 30. The information processing device 40 can perform information processing such as calculations using the input data. The information processing device 40 can transmit the results of the information processing to the information processing device 10 via the network 30. This can reduce the calculation load on the information processing device 10. The information processing device 10 can then transmit the data received from the information processing device 40 or data generated based on the data to the information terminal 20a via the network 31.

[0057] The information terminals 20a to 20d are information terminal devices such as personal computers (PCs) used by users, and may also be referred to as client PCs, client computers, etc. FIG. 1 illustrates, as examples, the information terminal 20a, which is a desktop computer; the information terminal 20b, which is a notebook computer; the information terminal 20c, which is a smartphone; and the information terminal 20d, which is a tablet computer. Note that the information terminal 20d can also be used as a notebook computer by connecting it to a housing 21 having a keyboard. The number of information terminals connected to the information processing device 10 is not particularly limited. While FIG. 1 shows four information terminals, the number of information terminals may be one, two, three, five, or more. Examples of such information terminals include desktop information terminals, notebook information terminals, tablet information terminals, mobile information terminals such as smartphones, and in-vehicle information terminals.

[0058] The information processing device 10 is a large computer such as a server computer or a supercomputer. The information processing device 10 is a computer with higher processing power than the information terminal 20. The information processing device 10 preferably has a function as a parallel computer. By using the information processing device 10 as a parallel computer, it is possible to perform large-scale calculations required for learning and inference of artificial intelligence (AI), for example.

[0059] The information processing device 10 can perform processing using a natural language processing model that uses AI. Examples of natural language processing models that use AI include BERT (Bidirectional Encoder Representations from Transformers) and T5 (Text-to-Text Transfer Transformer). The information processing device 10 may also be capable of performing processing using a model that uses a large-scale language model (such as a sentence generation model or a dialogue model).

[0060] The information processing device 10 can mainly execute processing using a specialized AI model that performs one or a small number of tasks. The specialized AI model is relatively small in scale and can be constructed within a local network. Specifically, the specialized AI model can be constructed within the network 31.

[0061] The information processing device 40 is a large computer such as a server computer or a supercomputer. The information processing device 40 is a computer with higher processing power than the information terminal 20. The information processing device 40 is also a computer with higher processing power than the information processing device 10. The information processing device 40 can also function as a parallel computer. By using the information processing device 40 as a parallel computer, it is possible to perform large-scale calculations required for AI learning and inference, for example. Note that when both the information processing device 10 and the information processing device 40 have the function of a parallel computer, the information processing device 40 has higher processing power and can perform large-scale calculations than the information processing device 10.

[0062] The information processing device 40 can perform processing using a dialogue model that utilizes a large-scale language model. Examples of large-scale language models include GPT-3, GPT-3.5, GPT-4, LaMDA (Language Model for Dialogue Applications), PaLM (Pathways Language Model), and PaLM2, and it is preferable to use GPT-4.

[0063] The information processing device 40 can execute processing using a general-purpose AI model that can perform various tasks. When the information processing device 10 and the information processing device 40 can each execute processing using a model that uses a large-scale language model, the information processing device 40 can execute processing using a larger-scale AI model than the information processing device 10.

[0064] Network 31 is a computer network smaller than network 30. Typically, network 30 is a global network, and network 31 is a local network. It is preferable to use the Internet, which is the foundation of the World Wide Web (WWW), as network 30. It is preferable to use an intranet or an extranet as network 31. In this embodiment, an example will be mainly described in which the Internet is used as network 30 and an intranet is used as network 31.

[0065] Other computer networks that can be used as the network 31 include a PAN (Personal Area Network), a LAN (Local Area Network), a CAN (Campus Area Network), a MAN (Metropolitan Area Network), a WAN (Wide Area Network), and a GAN (Global Area Network).

[0066] Furthermore, when wireless communication is performed, communication standards such as the fourth generation mobile communication system (4G), fifth generation mobile communication system (5G), and sixth generation mobile communication system (6G), or specifications standardized by the IEEE such as Wi-Fi (registered trademark) and Bluetooth (registered trademark), can be used as communication protocols or communication technologies.

[0067] The information processing system shown in FIG. 1 is suitable for a case where a user who uses the information terminal 20 and a service provider belong to the same organization such as a company.

[0068] For example, it is preferable that data be transmitted and received between the information terminal 20 and the information processing device 10 using a network 31 established within an organization such as a company. It is also preferable that data be transmitted and received between the information processing device 10 and the information processing device 40 using the Internet. This allows applications, tools, services, etc. provided outside the organization to be used more safely than when the information terminal 20 is directly connected to the information processing device 40 via the Internet. Furthermore, since processing using an AI model specialized for a specific application (the specialized AI model described above) can be performed within the organization by the information processing device 10, it is possible to assist users in entering data, thereby increasing user convenience and preventing the leakage of personal information, confidential information, etc. within the organization.

[0069] The flow of data in the information processing system shown in Fig. 1 will be described with reference to Fig. 2A. In Fig. 2A and other figures, the order of data transmission and reception is indicated by circled numbers.

[0070] First, the user inputs desired data DA1 to the information processing device 10 using the information terminal 20. The data DA1 is input from the information terminal 20 to the information processing device 10 via the network 31.

[0071] The data DA1 corresponds to a prompt that the user wants to input to the dialogue model. The format of the data DA1 is not particularly important, but it is preferably text data.

[0072] When the data DA1 includes data other than text data, it is preferable to convert the data DA1 into text data in the information processing device 10. When various data are converted into text data, it is preferable to use AI technology.

[0073] The information processing device 10 preferably has a function of generating text data using voice data. For example, the information processing device 10 can generate text data using voice data by performing processing using AI technology for voice recognition.

[0074] The information processing device 10 preferably has a function of generating text data using image data. For example, the information processing device 10 can generate text data using image data by performing processing using at least one AI technology, such as image recognition, handwritten character recognition, or caption generation. The information processing device 10 may also have an optical character recognition (OCR) function.

[0075] The information processing device 10 preferably has a function for generating text data using video data. For example, audio data and image data can be extracted from the video data, and text data can be generated from each of them using the above-mentioned methods. Furthermore, processing using AI technology for emotion analysis can be performed to read the speaker's facial expression and add additional information to the text data.

[0076] Furthermore, it is preferable that the information processing device 10 generates text data using brain measurement data. For example, it is preferable that functional magnetic resonance imaging (fMRI) be used to obtain brain-related measurement data such as electroencephalograms and brain activity, and that the text data be generated using the measurement data using AI technology. Note that the text data may be generated directly from the brain measurement data, or image data may be generated using the brain measurement data, and the text data may be generated using the image data.

[0077] Next, the information processing device 10 generates data DA2 using the data DA1. The following description will be given taking as an example a case where the data DA1 is text data. Note that if the data DA1 includes data other than text data, the data DA1 can be interpreted as text data generated using the data DA1.

[0078] The information processing device 10 extracts, from the data DA1, first text data corresponding to the instruction sentence and second text data corresponding to at least a part of the other sentence.

[0079] The instruction sentence is part of the prompt input to the dialogue model, and can be rephrased as a question sentence, an instruction sentence, etc. The second text data corresponds to a sentence other than the instruction sentence in the prompt, and includes an input sentence, a sentence explaining a condition, etc.

[0080] Specifically, the first text data may be an instruction such as "Under condition A, summarize input sentence B." The second text data may be "Condition A [any text]," "Input sentence B [any text]," etc.

[0081] The method for extracting the first text data and the second text data is not particularly limited, but it is preferable to use a natural language processing model, in particular an AI model specialized for the task of extracting sentences. For example, the first text data may be extracted from the data DA1 using an AI model, and other text data may be regarded as the second text data. Alternatively, for example, the first text data and the second text data may be extracted from the data DA1 using an AI model, and the remaining text data may be regarded as unnecessary text data and deleted.

[0082] Next, the information processing device 10 uses the first text data corresponding to the instruction sentence to obtain third text data corresponding to a modified sentence of the instruction sentence. The third text data is generated by modifying the first text data so that the instruction sentence becomes more appropriate for the dialogue model. This allows the instruction sentence included in the data DA1 to be modified to an expression that allows the dialogue model to return a more appropriate response. Compared to the first text data, the third text data can be said to be data that is expected to produce a more appropriate answer when input to the dialogue model.

[0083] If the first text data is "Summarize input sentence B under condition A," the third text data may be, for example, "Summarize input sentence B based on condition A."

[0084] The method for acquiring the third text data is not particularly limited. For example, the third text data may be acquired using a rule-based method. Alternatively, sentences may be vectorized using Bag of Words (BoW) or the like, and the third text data may be acquired using a classifier using a support vector machine (SVM) or a random forest (RF). It is preferable to use a natural language processing model, particularly an AI model specialized for the task of proofreading text. For example, it is preferable to use a natural language processing model that has been fine-tuned in advance using an instruction sentence before correction and an instruction sentence after correction. For example, if the information processing device 10 is capable of performing processing using a sentence generation model that utilizes a large-scale language model, the third text data can be acquired by inputting a prompt including the first text data into the sentence generation model. The first text data and the generated third text data can also be used for retraining the sentence generation model.

[0085] The extraction of the first text data and the second text data and the acquisition of the third text data may be performed simultaneously using a single AI model. For example, the information processing device 10 may perform processing using a sentence generation model that utilizes a large-scale language model. Specifically, by inputting a prompt including the data DA1 into the sentence generation model, the first text data can be extracted and the third text data generated using the first text data can be acquired. In this case, as described above, data other than the first text data in the data DA1 may be considered as the second text data, or the first text data and the second text data may be extracted from the data DA1, and the remaining text data may be considered as unnecessary text data and deleted.

[0086] The second text data and the third text data obtained through such processing are input as data DA2 from information processing device 10 to information processing device 40 via network 30.

[0087] The types and number of processes performed to generate data DA2 from data DA1 are not particularly limited. For example, the information processing device 10 may perform one or more of the following on data DA1: a process to correct typos and omissions, a process to determine whether data DA1 contains at least one of personal information, confidential information, and malicious instructions, and a process to delete at least one of personal information, confidential information, and malicious instructions. These processes may be performed on the second text data and the third text data after the third text data is generated. Alternatively, these processes may be performed on data DA1, and then the first text data and the second text data may be extracted and the third text data may be generated.

[0088] The method for correcting typos, omissions, etc. in the data DA1 is not particularly limited. For example, it is preferable to use a natural language processing model, particularly an AI model specialized for the task of proofreading text. For example, it is preferable to use a natural language processing model that has been fine-tuned in advance using instruction sentences before and after proofreading. Alternatively, for example, the information processing device 10 may perform processing using a sentence generation model that utilizes a large-scale language model. Specifically, by inputting into the sentence generation model a prompt that includes an instruction sentence for correcting typos, omissions, etc. in an input sentence and the data DA1 as the input sentence, it is possible to obtain text data in which typos, omissions, etc. in the data DA1 have been corrected.

[0089] The method for determining whether personal information is included in the data DA1 and the method for deleting the personal information included in the data DA1 are not particularly limited. For example, it is preferable to use a natural language processing model, particularly an AI model specialized for the task of determining whether personal information is included. For example, it is preferable to use a natural language processing model that has been fine-tuned using data annotated with personal information in advance. Alternatively, for example, the information processing device 10 may perform processing using a sentence generation model that utilizes a large-scale language model. Specifically, by inputting a prompt including an instruction statement for deleting personal information included in an input sentence and the data DA1 as the input sentence into the sentence generation model, it is possible to obtain text data from which personal information has been deleted from the data DA1.

[0090] The method for determining whether confidential information is included in the data DA1 and the method for deleting the confidential information included in the data DA1 are not particularly limited. For example, it is preferable to use a natural language processing model, particularly an AI model specialized for the task of determining whether confidential information is included. For example, it is preferable to use a natural language processing model that has been fine-tuned using data annotated with confidential information in advance. Furthermore, it is preferable to be able to update the AI ​​model as needed by updating the annotation data for the confidential information. This makes it possible to respond to confidential information that changes daily. Alternatively, for example, the information processing device 10 may perform processing using a sentence generation model that utilizes a large-scale language model. Specifically, by inputting a prompt into the sentence generation model that includes an instruction statement for deleting confidential information included in an input sentence and the data DA1 as the input sentence, it is possible to obtain text data from the data DA1 from which confidential information has been deleted. It is also preferable to use few-shot prompting. Specifically, it is preferable that the prompt include an explanation (such as a definition or example) of the confidential information, and that the explanation be updateable. This makes it possible to respond to confidential information that changes daily.

[0091] The method for determining whether the data DA1 contains malicious instructions and the method for deleting the malicious instructions contained in the data DA1 are not particularly limited. For example, it is preferable to use a natural language processing model, particularly an AI model specialized for the task of determining whether or not malicious instructions are present. For example, it is preferable to use a natural language processing model that has been fine-tuned using data annotated with malicious instructions in advance. Furthermore, it is preferable to be able to update the AI ​​model as needed by updating the annotation data for malicious instructions. This makes it possible to respond to prompt injection techniques, fraudulent commands, and the like that change daily. Alternatively, for example, the information processing device 10 may perform processing using a sentence generation model that utilizes a large-scale language model. Specifically, by inputting a prompt into the sentence generation model that includes an instruction statement for deleting the malicious instructions contained in the input sentence and the data DA1 as the input sentence, it is possible to obtain text data from the data DA1 from which the malicious instructions have been deleted. It is also preferable to use few-shot prompting. Specifically, it is preferable that the prompt include an explanation (definition, example, etc.) of the malicious instructions, and that the explanation be updatable. This makes it possible to respond to prompt injection techniques and fraudulent commands that change daily.

[0092] In this way, it is preferable that the second text data and the third text data are input as data DA2 from the information processing device 10 to the information processing device 40 via the network 30 after correcting typos and omissions and deleting at least one of personal information, confidential information, and malicious instructions.

[0093] Next, data DB1 is input from information processing device 40 to information processing device 10 via network 30. Data DB1 is text data equivalent to a response sentence corresponding to data DA2 input to information processing device 40.

[0094] Next, the information processing device 10 outputs the data DB2 to the information terminal 20 via the network 31. The data DB2 is data based on the data DB1 input to the information processing device 10. In this way, the user can obtain the data DB2 corresponding to the response sentence to the data DA1.

[0095] For example, data DB1 may be converted into data other than text data and output to information terminal 20 as data DB2. For example, if data DA1 is data other than text data, data DB2 is preferably data in the same format as data DA1. It is preferable that information processing device 10 has a function of generating audio data or image data using text data, for example. Note that data DB1 may be used as data DB2 as is.

[0096] Fig. 3 is a block diagram of the information processing system shown in Fig. 1. In Fig. 3, an example of the configuration of the information processing device 10 will be mainly described.

[0097] The information processing device 10 includes a receiving unit 110 , a storage unit 120 , a processing unit 130 , an output unit 140 , and a transmission path 150 .

[0098] In the drawings accompanying this specification, the components are classified by function and shown as independent blocks in the block diagrams. However, in reality, it is difficult to completely separate the components by function, and one component may be involved in multiple functions. For example, part of the processing unit 130 may function as the reception unit 110. Also, one function may be involved in multiple components. For example, the processing performed by the processing unit 130 may be executed by different servers depending on the processing.

[0099] [Reception Unit 110] The reception unit 110 can receive data from outside the information processing device 10.

[0100] The receiving unit 110 receives data from the information terminal 20 via the network 31. As described above, the data is preferably text data. If the data is data other than text data, the processing unit 130 preferably generates text data using the data received by the receiving unit 110.

[0101] The receiving unit 110 may also receive a user's evaluation of the data output by the information processing device 10 to the information terminal 20 .

[0102] The receiving unit 110 also receives data from the information processing device 40 via the network 30 .

[0103] The information supplied to the reception unit 110 is supplied to one or both of the storage unit 120 and the processing unit 130 via the transmission path 150 .

[0104] [Storage Unit 120] The storage unit 120 has a function of storing a program executed by the processing unit 130. The storage unit 120 may also have a function of storing data generated by the processing unit 130 (e.g., calculation results, analysis results, inference results), data input to the reception unit 110, etc.

[0105] The storage unit 120 may have a database. Furthermore, the information processing device 10 may have a database separate from the storage unit 120. The information processing device 10 may have a function to retrieve data from a database that exists outside the storage unit 120, outside the information processing device 10, or outside the information processing system. Furthermore, the information processing device 10 may have a function to retrieve data from both its own database and an external database.

[0106] A file server may be used instead of the database. For example, when using files stored in a file server, it is preferable that the database has paths to files stored in the file server.

[0107] The storage unit 120 includes at least one of a volatile memory and a non-volatile memory. Examples of the volatile memory include a dynamic random access memory (DRAM) and a static random access memory (SRAM). Examples of the non-volatile memory include a resistive random access memory (ReRAM), a phase change random access memory (PRAM), a ferroelectric random access memory (FeRAM), a magnetoresistive random access memory (MRAM), and a flash memory. The storage unit 120 may also include a recording media drive. Examples of the recording media drive include a hard disk drive (HDD) and a solid state drive (SSD). Various types of storage may be used for the storage unit 120.

[0108] [Processing Unit 130] The processing unit 130 has a function of performing processes such as calculation, analysis, and inference using data supplied from one or both of the receiving unit 110 and the storage unit 120. The processing unit 130 can supply the generated data (e.g., calculation results, analysis results, and inference results) to one or both of the storage unit 120 and the output unit 140.

[0109] The processing unit 130 has a function of acquiring data from the storage unit 120 .

[0110] The processing unit 130 has a function of using the data received by the receiving unit 110 to obtain first text data corresponding to the instruction sentence and second text data corresponding to at least a part of the other sentence.

[0111] The processing unit 130 also has a function of inputting the first text data into the natural language processing model to obtain third text data corresponding to a corrected sentence of the instruction sentence.

[0112] Furthermore, it is preferable that the processing unit 130 has a function of inputting data into the determination model to obtain a determination result as to whether or not data can be input into the dialogue model. Here, the data input into the determination model may be the data itself accepted by the accepting unit 110, or may be the second text data and the third text data.

[0113] The judgment model preferably makes a judgment based on the presence or absence of at least one of personal information, confidential information, and malicious instructions.

[0114] In addition, it is preferable that the processing unit 130 has a function of acquiring third data when the determination result by the determination model is negative, that is, when it is determined that the data contains any one of personal information, confidential information, and malicious instructions. The third data is data that is generated using the data input to the determination model and for which the determination result of whether the data can be input to the dialogue model is positive.

[0115] For example, when the data itself received by the receiving unit 110 is input to the determination model, it is preferable to use the generated third data to obtain the first text data and the second text data. Also, when the second text data and the third text data are input to the determination model, the generated third data can be used as data to be input to the dialogue model, i.e., data to be input to the information processing device 40.

[0116] The processing unit 130 may include, for example, an arithmetic circuit, a central processing unit (CPU), and a graphics processing unit (GPU).

[0117] The processing unit 130 may have a microprocessor such as a DSP (Digital Signal Processor). The microprocessor may be implemented by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array) or an FPAA (Field Programmable Analog Array). The processing unit 130 may also have a quantum processor. The processing unit 130 can perform various data processing and program control by interpreting and executing instructions from various programs using the processor. Programs that can be executed by the processor are stored in at least one of the memory area of ​​the processor and the storage unit 120.

[0118] The processing unit 130 may have a main memory, which may include at least one of a volatile memory such as a random access memory (RAM) and a non-volatile memory such as a read only memory (ROM).

[0119] The RAM may be, for example, a DRAM or an SRAM, and a virtual memory space is allocated and used as a working space for the processing unit 130. The operating system, application programs, program modules, program data, lookup tables, and the like stored in the storage unit 120 are loaded into the RAM for execution. The data, programs, and program modules loaded into the RAM are each directly accessed and operated by the processing unit 130.

[0120] The ROM can store a BIOS (Basic Input / Output System), firmware, etc., which do not require rewriting. Examples of ROM include mask ROM, OTPROM (One Time Programmable Read Only Memory), and EPROM (Erasable Programmable Read Only Memory). Examples of EPROMs include UV-EPROMs (Ultra-Violet Erasable Programmable Read Only Memories), which allow stored data to be erased by exposure to ultraviolet light, EEPROMs (Electrically Erasable Programmable Read Only Memories), and flash memories.

[0121] It is preferable that the information processing device 10 uses AI for at least some of its processing.

[0122] In particular, it is preferable that the information processing device 10 uses an artificial neural network (ANN, hereinafter also simply referred to as a neural network). A neural network is realized by a circuit (hardware) or a program (software).

[0123] In this specification, a neural network refers to a general model that mimics the neural circuit network of a living organism, determines the connection strength between neurons through learning, and has problem-solving capabilities. A neural network has an input layer, an intermediate layer (hidden layer), and an output layer.

[0124] In this specification and the like, when discussing neural networks, determining the connection strengths (also called weighting coefficients) between neurons from existing information may be referred to as "learning."

[0125] In this specification and the like, the act of constructing a neural network using connection strengths obtained by learning and deriving a new conclusion from it may be referred to as "inference."

[0126] [Output Unit 140] The output unit 140 outputs information based on the processing result of the processing unit 130. The output unit 140 can supply at least one of the calculation result, analysis result, and inference result of the processing unit 130 to an external device of the information processing device 10.

[0127] The output unit 140 outputs data to the information terminal 20 via the network 31. The output unit 140 can output the data itself received from the information processing device 40 to the information terminal 20. Alternatively, the output unit 140 can output data generated by the processing unit 130 based on the data received from the information processing device 40 to the information terminal 20.

[0128] The information can be presented, for example, by displaying it on the display screen of the information terminal 20 and / or by outputting a file in CSV format or the like.

[0129] Furthermore, the output unit 140 outputs data to the information processing device 10 via the network 30 based on the processing result of the processing unit 130. The output unit 140 does not output the data received from the information terminal 20 directly to the information processing device 10, but outputs data generated by the processing unit 130 based on the data to the information processing device 40.

[0130] [Transmission Path 150] The transmission path 150 has a function of transmitting data. Data can be transmitted and received between the reception unit 110, the storage unit 120, the processing unit 130, and the output unit 140 via the transmission path 150.

[0131] <Configuration Example 2 of Information Processing System> Fig. 4 shows an example of an information processing system according to this embodiment. In the information processing system shown in Fig. 4, an information processing device 11 is connected to an information processing device 40 via a network 30. The information processing device 11 is also connected to an information processing device 15 and a plurality of information terminals 20 via a network 31. In Fig. 4, information terminals 20a and 20b are shown as examples of the information terminals 20.

[0132] The flow of data in the information processing system shown in FIG. 4 will be described with reference to FIG. 5A.

[0133] 4 is mainly used to relay transmission and reception of data between the information terminal 20 and the information processing device 40. Then, in the information processing device 15, data DA2 is generated using the data DA1.

[0134] First, the user inputs desired data DA1 to the information processing device 11 using the information terminal 20. The data DA1 is input from the information terminal 20 to the information processing device 11 via the network 31.

[0135] Next, the information processing device 11 inputs the data DA1 to the information processing device 15 via the network 31.

[0136] Next, the information processing device 15 generates data DA2 using the data DA1.

[0137] The information processing device 15 is configured to execute the various processes performed by the information processing device 10. Specifically, the information processing device 15 extracts, from the data DA1, first text data corresponding to the instruction sentence and second text data corresponding to at least a part of the other sentences. Then, the information processing device 15 uses the first text data to obtain third text data corresponding to a correction sentence of the instruction sentence.

[0138] Furthermore, it is preferable that the information processing device 15 has a function of converting the data DA1 into text data.

[0139] In addition, the information processing device 15 may perform one or more of the following processes on the data DA1: correcting typos and omissions, determining whether the data DA1 contains at least one of personal information, confidential information, and malicious instructions, and deleting at least one of personal information, confidential information, and malicious instructions.

[0140] Next, the information processing device 15 inputs the data DA2 to the information processing device 11 via the network 31.

[0141] Next, the information processing device 11 inputs the data DA2 to the information processing device 40 via the network 30.

[0142] Next, data DB1 is input from information processing device 40 to information processing device 11 via network 30. Data DB1 is text data equivalent to a response sentence corresponding to data DA2 input to information processing device 40.

[0143] Next, the information processing device 11 outputs the data DB2 to the information terminal 20 via the network 31. The data DB2 is data based on the data DB1 input to the information processing device 11. In this way, the user can obtain the data DB2 corresponding to the response sentence to the data DA1.

[0144] The data DB2 can be generated by the information processing device 11 or the information processing device 15. For example, the information processing device 11 may input the data DB1 input from the information processing device 40 to the information processing device 15. Then, the information processing device 15 may generate the data DB2 based on the data DB1.

[0145] The information processing system shown in FIG. 5A is suitable for a case where a user who uses information terminal 20 and a service provider belong to the same organization such as a company, similar to the information processing system shown in FIG.

[0146] For example, data is preferably transmitted and received between the information terminal 20 and the information processing device 11 using a network 31 established within an organization such as a company. Data is also preferably transmitted and received between the information processing device 15 and the information processing device 11 using a network 31 established within an organization such as a company. Data is also preferably transmitted and received between the information processing device 11 and the information processing device 40 using the Internet. This allows applications, tools, services, etc. provided outside the organization to be used more safely than when the information terminal 20 is directly connected to the information processing device 40 via the Internet. Furthermore, since processing using an AI model specialized for the application can be performed within the organization by the information processing device 15, data input by the user can be supported, improving user convenience while preventing the leakage of personal information, confidential information, etc. within the organization. However, connecting the information processing device 15 to the Internet raises the risk of external cyberattacks, etc. The information processing system shown in FIG. 5 does not require the information processing device 15 to be directly connected to the information processing device 40 via the Internet, which is preferable from a security perspective.

[0147] Fig. 6 is a block diagram of the information processing system shown in Fig. 4. In Fig. 6, an example of the configuration of the information processing device 11 and the information processing device 15 will be mainly described.

[0148] The information processing device 11 includes a receiving unit 110 , a storage unit 120 , a processing unit 130 , an output unit 140 , and a transmission path 150 .

[0149] For the reception unit 110, storage unit 120, output unit 140, and transmission path 150 of the information processing device 11, the descriptions of the information processing device 10 can be referred to. The following mainly describes the differences from the information processing device 10.

[0150] The information processing device 15 includes a receiving unit 115 , a storage unit 125 , a processing unit 135 , an output unit 145 , and a transmission path 155 .

[0151] The receiving unit 110 receives data from the output unit 145 via the network 31 .

[0152] The receiving unit 115 receives data from the output unit 140 via the network 31 .

[0153] The information supplied to the reception unit 115 is supplied to one or both of the storage unit 125 and the processing unit 135 via the transmission path 155 .

[0154] The storage unit 125 has a function of storing a program executed by the processing unit 135. The storage unit 125 may also have a function of storing data generated by the processing unit 135 (e.g., calculation results, analysis results, inference results), data input to the reception unit 115, etc. The storage unit 125 may have various configurations applicable to the storage unit 120.

[0155] The processing unit 135 has a function of performing processes such as calculation, analysis, and inference using data supplied from one or both of the reception unit 115 and the storage unit 125. The processing unit 135 can supply generated data (e.g., calculation results, analysis results, and inference results) to one or both of the storage unit 125 and the output unit 145. The processing unit 135 can use various configurations applicable to the processing unit 130.

[0156] The processing unit 135 has a function of acquiring data from the storage unit 125 .

[0157] The processing unit 135 has a function of using the data received by the receiving unit 115 to obtain first text data corresponding to the instruction sentence and second text data corresponding to at least a part of the other sentence.

[0158] The processing unit 135 also has a function of inputting the first text data into the natural language processing model to obtain third text data corresponding to a corrected sentence of the instruction sentence.

[0159] Furthermore, it is preferable that the processing unit 135 has a function of inputting data into the determination model to obtain a determination result as to whether or not data can be input into the dialogue model. Here, the data input into the determination model may be the data itself accepted by the accepting unit 115, or may be the second text data and the third text data.

[0160] The judgment model preferably makes a judgment based on the presence or absence of at least one of personal information, confidential information, and malicious instructions.

[0161] In addition, it is preferable that the processing unit 135 has a function of acquiring third data when the determination result by the determination model is negative, that is, when it is determined that the data contains any one of personal information, confidential information, and malicious instructions. The third data is data that is generated using the data input to the determination model and for which the determination result of whether the data can be input to the dialogue model is positive.

[0162] For example, when the data itself accepted by the accepting unit 115 is input to the determination model, it is preferable to use the generated third data to acquire the first text data and the second text data. Also, when the second text data and the third text data are input to the determination model, the generated third data can be used as data to be input to the dialogue model, i.e., data to be input to the information processing device 40.

[0163] The information processing device 15 preferably uses AI for at least a part of the processing, and preferably uses a neural network.

[0164] The output unit 140 outputs the data to the reception unit 115 via the network 31 .

[0165] The output unit 145 outputs information based on the processing result of the processing unit 135. The output unit 145 can supply at least one of the calculation result, analysis result, and inference result of the processing unit 135 to an external device of the information processing device 15.

[0166] The output unit 145 outputs data to the reception unit 110 of the information processing device 10 via the network 31 .

[0167] The transmission path 155 has a function of transmitting data. Data can be transmitted and received between the reception unit 115, the storage unit 125, the processing unit 135, and the output unit 145 via the transmission path 155.

[0168] The processing section 130 and the processing section 135 can each include one or both of a transistor having a metal oxide in a channel formation region (OS transistor) and a transistor having silicon in a channel formation region (Si transistor).

[0169] Note that in this specification and the like, a transistor whose channel formation region includes an oxide semiconductor or a metal oxide is referred to as an oxide semiconductor transistor or an OS transistor. The channel formation region of an OS transistor preferably includes a metal oxide.

[0170] In this specification and the like, a metal oxide refers to an oxide of a metal in a broad sense. Metal oxides are classified into oxide insulators, oxide conductors (including transparent oxide conductors), oxide semiconductors (also referred to as oxide semiconductors or simply as OSs), and the like. For example, when a metal oxide is used for a semiconductor layer of a transistor, the metal oxide may be referred to as an oxide semiconductor.

[0171] The metal oxide included in the channel formation region preferably contains indium (In). When the metal oxide included in the channel formation region contains indium, the carrier mobility (electron mobility) of the OS transistor is increased. Furthermore, the metal oxide included in the channel formation region is preferably an oxide semiconductor containing element M. The element M is preferably at least one of aluminum (Al), gallium (Ga), and tin (Sn). Other elements applicable to element M include boron (B), silicon (Si), titanium (Ti), iron (Fe), nickel (Ni), germanium (Ge), yttrium (Y), zirconium (Zr), molybdenum (Mo), lanthanum (La), cerium (Ce), neodymium (Nd), hafnium (Hf), tantalum (Ta), and tungsten (W). However, a combination of two or more of the above elements may be used as element M. The element M is, for example, an element having a high bond energy with oxygen. For example, it is an element having a higher bond energy with oxygen than indium. The metal oxide contained in the channel formation region is preferably a metal oxide containing zinc (Zn), since zinc-containing metal oxides may be easily crystallized.

[0172] The metal oxide contained in the channel formation region is not limited to a metal oxide containing indium. The semiconductor layer may be, for example, a metal oxide containing zinc but not indium, such as zinc tin oxide or gallium tin oxide, a metal oxide containing gallium, or a metal oxide containing tin.

[0173] The processing unit 130 and the processing unit 135 each preferably include an OS transistor. Because an OS transistor has an extremely low off-state current, the OS transistor can be used as a switch for retaining charge (data) flowing into a capacitor functioning as a memory element, thereby ensuring a long data retention period. By using this characteristic in at least one of the register and cache memory of the processing unit, the processing unit can be operated only when necessary, and can be turned off in other cases by saving information from the previous processing in the memory element. In other words, normally-off computing is possible, and the power consumption of the information processing system can be reduced.

[0174] <Configuration Example 3 of Information Processing System> FIG. 7 shows an example of an information processing system according to the present embodiment. In the information processing system shown in FIG. 7, an information processing device 10 is connected to an information processing device 40 and multiple information terminals 20 via a network 30. In FIG. 7, in addition to the information terminals 20a to 20d described above, an in-vehicle information terminal 20e is shown as an example of the information terminal 20. As shown in FIG. 7, the information processing system according to the present embodiment can be used for various purposes. For example, although not shown, the information processing system according to the present embodiment can also be applied to information terminals with touch panels installed in stores and the like that are part of a table ordering system, information terminals installed in convenience stores, and the like. This makes it possible to provide an information processing system that is easy for users to use.

[0175] The flow of data in the information processing system shown in Fig. 7 is as shown in Fig. 2B. Fig. 2B is the same as Fig. 2A except that data transmission and reception between the information terminal 20 and the information processing device 10 is performed via the network 30, and therefore a detailed description thereof will be omitted.

[0176] A user of the information processing system can access the information processing device 10 from an information terminal 20. The user can then receive a service using the information processing method according to one aspect of the present invention.

[0177] The information processing system shown in FIG. 7 is suitable for a case where a user who uses the information terminal 20 and a service provider belong to different organizations such as companies.

[0178] The information processing device 10 is connected to an information terminal 20 used by a user receiving a service via a network 30 (typically, the Internet). A service provider can provide the service using the information processing device 10. Furthermore, data is transmitted and received between the information processing device 10 and the information processing device 40 using the network 30 (typically, the Internet). This allows applications, tools, services, etc. using the information processing device 40 to be used more safely than when the information terminal 20 is directly connected to the information processing device 40 via the Internet. Furthermore, by performing processing using an AI model specialized for the application on the information processing device 10, it is possible to assist the user in entering data, thereby increasing user convenience and preventing the leakage of personal information, confidential information, etc. within an organization. In other words, it is possible to provide a service that is beneficial to the user.

[0179] <Configuration Example 4 of Information Processing System> Fig. 8 shows an example of an information processing system according to this embodiment. In the information processing system shown in Fig. 8, an information processing device 11 is connected to an information processing device 40 and a plurality of information terminals 20 via a network 30. In Fig. 8, the above-mentioned information terminals 20a to 20e are shown as examples of the information terminals 20. Furthermore, the information processing device 11 is connected to an information processing device 15 via a network 31.

[0180] The flow of data in the information processing system shown in Fig. 8 is as shown in Fig. 5B. Fig. 5B is the same as Fig. 5A except that data transmission and reception between the information terminal 20 and the information processing device 10 is performed via the network 30, and therefore a detailed description thereof will be omitted.

[0181] A user of the information processing system can access the information processing device 11 from an information terminal 20. The user can then receive services using the information processing method according to one aspect of the present invention.

[0182] The information processing system shown in FIG. 8, like the information system shown in FIG. 7, is suitable for a case where a user who uses information terminal 20 and a service provider belong to different organizations such as companies.

[0183] The information processing device 11 is connected to an information terminal 20 used by a user receiving a service via the Internet. Preferably, data is exchanged between the information processing device 15 and the information processing device 11 using a network 31 established within an organization, such as a company, that provides the service. The service provider can provide the service using the information processing device 11 and the information processing device 15. Data is also exchanged between the information processing device 11 and the information processing device 40 via the Internet. This allows applications, tools, services, etc. using the information processing device 40 to be used more safely than when the information terminal 20 is directly connected to the information processing device 40 via the Internet. Furthermore, by performing processing using an AI model specialized for the application on the information processing device 15, data input by the user can be assisted, improving user convenience while preventing the leakage of personal information, confidential information, etc. within the organization. In other words, a service beneficial to the user can be provided. Furthermore, the information processing device 15 does not need to be directly connected to the information processing device 40 via the Internet, which is preferable from a security perspective.

[0184] A plurality of configuration examples shown in this embodiment mode can be combined as appropriate.

[0185] 10: Information processing device, 11: Information processing device, 15: Information processing device, 20a: Information terminal, 20b: Information terminal, 20c: Information terminal, 20d: Information terminal, 20e: Information terminal, 20: Information terminal, 21: Housing, 30: Network, 31: Network, 40: Information processing device, 110: Reception unit, 115: Reception unit, 120: Storage unit, 125: Storage unit, 130: Processing unit, 135: Processing unit, 140: Output unit, 145: Output unit, 150: Transmission path, 155: Transmission path

Claims

A receiving unit, a processing unit, and an output unit, the receiving unit receives first data from a computer via a first network; The processing unit performs a process of acquiring, using the first data, first text data corresponding to an instruction sentence and second text data corresponding to at least a part of another sentence; A process of acquiring third text data corresponding to a modified sentence of the instruction sentence by inputting the first text data into a natural language processing model; a process of acquiring fourth text data corresponding to a response sentence by inputting the second text data and the third text data into a dialogue model via a second network; The output unit outputs second data based on the fourth text data to the computer via the first network. 　 In claim 1, the first network is a local network; The information processing device, wherein the second network is a global network. 　 In claim 1, The information processing device, wherein the first network and the second network are each a global network. 　 In claim 1, the processing unit inputs the first data into a judgment model to obtain a judgment result as to whether the first data can be input into the dialogue model, and further, when the judgment result is no, obtains third data; the determination result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as a criterion; the third data is generated using the first data, and a determination result of whether or not input to the dialogue model is possible is possible; and The information processing device, wherein the first text data and the second text data are acquired using the third data. 　 In claim 1, The output unit outputs fourth data based on the third text data to the computer. 　 In claim 1, An information processing device, wherein the third text data is generated by modifying the first text data so that the first text data becomes an instruction sentence more appropriate for the dialogue model. 　 receiving first data from a computer via a first network; using the first data, to obtain first text data corresponding to the instruction sentence and second text data corresponding to at least a part of another sentence; inputting the first text data into a natural language processing model to obtain third text data corresponding to a modified sentence of the instruction sentence; acquiring fourth text data corresponding to a response sentence by inputting the second text data and the third text data into a dialogue model via a second network; outputting second data based on the fourth text data to the computer via the first network. 　 In claim 7, the first network is a local network; The information processing method, wherein the second network is a global network. 　 In claim 7, transmitting and receiving data to and from a processing unit capable of executing processing using the natural language processing model via a third network; the first network and the third network are local networks, The information processing method, wherein the second network is a global network. 　 In claim 7, transmitting and receiving data to and from a processing unit capable of executing processing using the natural language processing model via a third network; the first network and the second network are each a global network; The information processing method, wherein the third network is a local network. 　 In claim 7, inputting the first data into a judgment model to obtain a judgment result as to whether the first data can be input into the dialogue model, and further, if the judgment result is no, obtaining third data; the determination result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as a criterion; the third data is generated using the first data, and a determination result of whether or not input to the dialogue model is possible is possible; and The information processing method, wherein the first text data and the second text data are obtained using the third data. 　 In claim 7, and outputting fourth data based on the third text data to the computer. 　 a first computer, a second computer, and a third computer; the second computer receives first data from the first computer via a first network; the second computer obtains, using the first data, first text data corresponding to the instruction sentence and second text data corresponding to at least a part of another sentence; the second computer supplies the first text data to a natural language processing model to obtain third text data corresponding to a corrected sentence of the instruction sentence; the third computer receives the second text data and the third text data from the second computer via a second network; the third computer uses the second text data and the third text data as input data and performs processing using a dialogue model to generate fourth text data corresponding to a response sentence; the third computer outputs the fourth text data to the second computer via the second network; The second computer outputs second data based on the fourth text data to the first computer via the first network. 　 In claim 13, the first network is a local network; An information processing system, wherein the second network is a global network. 　 In claim 13, An information processing system, wherein the first network and the second network are each a global network. 　 In claim 13, the second computer inputs the first data into a judgment model to obtain a judgment result as to whether the first data can be input into the dialogue model, and further, if the judgment result is no, obtains third data; the determination result is a result determined using the presence or absence of at least one of personal information, confidential information, and malicious instructions as a criterion; the third data is generated using the first data, and a determination result of whether or not input to the dialogue model is possible is possible; and An information processing system, wherein the first text data and the second text data are obtained using the third data. 　 In claim 13, The second computer outputs fourth data based on the third text data to the first computer. 　 A first computer and a second computer, the first computer includes a receiving unit, a first processing unit, and an output unit; the second computer has a second processing unit; the receiving unit receives first data from a third computer via a first network; the first computer and the second computer are configured to transmit and receive data via a second network; the second processing unit uses the first data to obtain first text data corresponding to an instruction sentence and second text data corresponding to at least a part of another sentence; a process of acquiring third text data corresponding to a correction of the instruction sentence by inputting the first text data into a natural language processing model; the first processing unit is configured to execute a process of acquiring fourth text data corresponding to a response sentence by inputting the second text data and the third text data to a dialogue model via a third network; The output unit outputs second data based on the fourth text data to the third computer via the first network. 　 In claim 18, the first network and the second network are each a local network; An information processing system, wherein the third network is a global network. 　 In claim 18, the first network and the third network are each a global network; An information processing system, wherein the second network is a local network.

Images