Information providing device, information providing method, and information providing program

Abstract

To increase the accuracy of an answer output from a model.SOLUTION: An information providing device according to an embodiment has a collection unit and a generation unit. The collection unit collects a data source satisfying a condition related to reliability and provided from a providing source. The generation unit generates a language model that is generated by learning the collected data source, and that generates an answer to a prompt input by a user. The data source includes distribution data distributed by the distribution source to a distribution target user, and comment data to the distribution data posted by the distribution target user.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The disclosed embodiments relate to an information providing device, an information providing method, and an information providing program. [Background technology]

[0002] 2. Description of the Related Art Conventionally, there is known a system that uses a generative model to generate an answer to a question input by a user (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2022-503838 Summary of the Invention [Problem to be solved by the invention]

[0004] Conventional techniques leave room for improvement in terms of increasing the accuracy of answers output by models.

[0005] The present invention has been made in view of the above, and has an object to improve the accuracy of answers output from a model. [Means for solving the problem]

[0006] According to an embodiment, an information providing device includes a collection unit that collects data sources provided by a provider that satisfies a reliability condition, and a generation unit that generates a language model by learning the collected data sources, the language model generating an answer to a prompt input by a user. The data sources include distribution data distributed by the provider to a distribution target user and comment data posted by the distribution target user on the distribution data. Effect of the Invention

[0007] According to one aspect of the embodiment, the accuracy of the answers output by the model can be improved. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating an overview of an information providing device according to an embodiment. [Diagram 2] FIG. 2 is a functional block diagram illustrating an example of the configuration of an information providing device according to the embodiment. [Diagram 3] FIG. 3 is a diagram for explaining changes in the amount of data. [Figure 4] FIG. 4 is a flowchart illustrating the flow of the reliability determination process. [Diagram 5] FIG. 5 is a flowchart illustrating the flow of the pre-training process. [Figure 6] FIG. 6 is a flowchart illustrating the flow of the information provision process. [Figure 7] FIG. 7 is a diagram illustrating an example of a computer hardware configuration that functions as an information providing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] The present invention will be described below through embodiments, but the following embodiments do not limit the scope of the invention according to the claims. In addition, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0010] A process flow of the information providing device according to the embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram for explaining an overview of the information providing device according to the embodiment. Fig. 1 shows a configuration of an information providing system 1 including an information providing device 10 according to the embodiment.

[0011] As shown in Fig. 1, the information providing system 1 includes an information providing device 10, a data source provider 20, and a response destination 30. The data source provider 20 is, in substance, a server 20-1 for distributing information, a user terminal 20-2 owned by a user, a vehicle 20-3 with an automatic driving function, etc. The server 20-1 is, for example, a server managed by a newspaper company, a news agency, a broadcasting company, a publisher, an online news provider, etc. The server 20-1 distributes information via a website or a social networking service (SNS) on the Internet.

[0012] The answer destination 30 is a destination to which an answer to a user's question is provided using a language model included in the information providing device 10. The answer destination 30 is actually a user terminal 30-1, a vehicle 30-2 having an automatic driving function, or the like.

[0013] 1, the information providing device 10 acquires information about the provider from the data source provider 20 (step S1). The information about the provider is attribute information about the provider (business content, business scale, number of employees, number of registered members), and information about the area where the business is conducted.

[0014] Next, the information providing device 10 judges whether the destination satisfies the condition regarding reliability based on the information on the source (step S2). For example, the information providing device 10 judges that the source (newspaper company, news agency, broadcasting company, publisher, net news provider) whose business is news distribution satisfies the condition regarding reliability. In other words, the information providing device 10 judges that the data source of news articles provided by the source whose business is news distribution satisfies the condition regarding reliability because the reliability is high. In addition, the information providing device 10 judges that the source satisfies the condition regarding reliability with a business scale of a certain level or more. The business scale is, for example, the number of news distributions (number of copies in the case of a newspaper company, number of users to whom the news agency has distributed, viewer rating in the case of a broadcasting company, number of copies in the case of a publisher, number of distributions or number of registered users in the case of a net news provider). In other words, the information providing device 10 judges that the source whose news articles are distributed to a certain level or more of users satisfies the condition regarding reliability.

[0015] Furthermore, when the data source provider 20 is a device such as a user terminal 20-2 or a vehicle 20-3, the information providing device 10 determines whether or not the manufacturer of the device satisfies a condition regarding reliability. For example, the information providing device 10 determines that a manufacturer with sales amount equal to or greater than a certain amount satisfies the condition regarding reliability.

[0016] Next, the information providing device 10 permits communication connection to the data source provider 20 that satisfies the reliability condition, and collects data sources from the data source provider 20 (step S3). The information providing device 10 and the data source provider 20 are connected for communication via a VPN (Virtual Private Network).

[0017] The data source is distribution data in the form of text, voice, image, etc. distributed by the server 20-1 of the data source provider 20. The data source is also user posted content (text, voice, image, etc.) provided by the user terminal 20-2. The data source is past travel route provided by the vehicle 20-3, congestion status of roads and sidewalks during travel, travel time required to reach the destination, behavior of the vehicle 20-3 when traveling to the destination (accelerator, brake, steering, etc.), etc. The data source also includes comment data posted by a user (distribution target user) in response to the distribution data. The comment data is, for example, a user's comment on a news article (distribution data).

[0018] The information providing device 10 learns a model based on the acquired data source. The model is a language model. For example, ChatGPT by OpenAI is known as a language model (Reference: https: / / openai.com / blog / chatgpt). The language model may be a generative model such as GAN (Generative Adversarial Networks) or VAE (Variational Autoencoder) using a neural network.

[0019] In this embodiment, the language model generates a text response to a prompt input by a user. The model is pre-trained using a data source. The model may be trained by known machine learning techniques.

[0020] This allows the model to generate answers based on the data source. For example, if the prompt contains the keyword "electric car," the model can generate answers based on news articles related to that keyword. In other words, the content of the news articles influences how the model generates answers.

[0021] In step S1, the information providing device 10 continuously acquires data sources. Therefore, the amount of data sources acquired by the information providing device 10 increases over time.

[0022] The information providing device 10 has two models that use different data sources for learning: a standard model and an advanced model.

[0023] The information providing device 10 learns two language models based on the data sources (step S4). Specifically, the information providing device 10 learns a standard model based on the past data sources. The information providing device 10 also learns an advanced model based on the current data sources.

[0024] Here, the past data source is a data source acquired up to a certain time period in the past before the timing of model learning. For example, if the certain period is two years, when learning is performed at "2023 / 6 / 22 16:05", the past data source is a data source acquired from the start of data source acquisition up to "2021 / 6 / 22 16:05".

[0025] On the other hand, the current data source is the data source acquired up to the time when the model is trained. For example, if training is performed on "2023 / 6 / 22 16:05", the current data source is the data source acquired from the start of data source acquisition up to "2023 / 6 / 22 16:05".

[0026] The timing for learning the model may be every second, or may be every shorter time (for example, every nanosecond).

[0027] In addition, the information providing device 10 uses the delivery data and comment data included in the data source for model learning. This comment data may include supplementary information to the delivery data, more detailed information, information from a different perspective, etc., so by using such comment data for learning, it is possible to further improve the accuracy of answers output from the model.

[0028] Furthermore, the information providing device 10 learns a model based on comments with positive content (positive comments) and comments with negative content (negative comments) in the comment data. Comments with positive content are comments that praise, approve, affirm, or supplement the distributed data. Comments with negative content are comments that criticize, oppose, or deny the distributed data, comments that contain extreme language, and comments with discriminatory content.

[0029] For example, the information providing device 10 prohibits negative comments from being used in model learning among the comment data. This prevents a response based on a negative comment from being output from the model, thereby reducing discomfort felt by a user who receives a response.

[0030] Furthermore, when the number of comments in the comment data is equal to or greater than a threshold, the information providing device 10 uses the comment data for model learning. That is, the information providing device 10 determines whether the number of comments in the comment data for each distribution data is equal to or greater than a threshold, and when it is equal to or greater than the threshold, uses the distribution data and the comment data for model learning, and when the number of comments is less than the threshold, uses only the distribution data for learning. This makes it possible to avoid a situation where the diversity of comments cannot be ensured due to a small number of comments.

[0031] Furthermore, the information providing device 10 adjusts the amount of learning for the model based on the number of positive comments and the number of negative comments in the number of comments. For example, when the ratio of negative comments among all comments in the comment data is equal to or greater than a threshold, the information providing device 10 prohibits the use of all comments or negative comments for learning the model.

[0032] Furthermore, the information providing device 10 increases the learning amount of distribution data in which the ratio of positive comments is higher than the ratio of negative comments among all comments in the comment data, compared to distribution data in which the ratio of positive comments is lower than the ratio of negative comments. For example, in the case of distribution data in which the ratio of positive comments is lower than the ratio of negative comments, the information providing device 10 prevents content related to the negative comments from being used in learning the model.

[0033] In addition, the information providing device 10 excludes sources that have a predetermined number or more of distribution data with negative comments exceeding a threshold from the collection of data sources. This makes it possible to exclude sources that distribute a lot of distribution data with a lot of negative comments, in other words, sources that distribute unreliable distribution data, from the collection of data sources, thereby increasing the reliability of the collected data sources.

[0034] Here, the information providing device 10 receives a prompt from the user as an input to the language model (step S5), and generates an answer to the prompt using a model according to the user's plan (step S6).

[0035] The prompt from the user is transmitted to the information providing device 10 via the answer providing destination 30. In addition, the information providing device 10 transmits the answer to the answer providing destination 30.

[0036] The information providing device 10 provides answers as a service. The service includes a higher-level premium plan and a lower-level freemium plan. The premium plan is a subscription plan that charges a fixed fee for a certain period (e.g., yearly or monthly), or a pay-as-you-go plan where the fee is determined according to the usage fee. The freemium plan is a free plan. The freemium plan may be replaced with a plan (e.g., a basic plan) that charges less than the premium plan.

[0037] The information providing device 10 realizes differentiation of service quality for each plan by the difference in data used for learning the model to be used. That is, when the user subscribes to the premium plan, the information providing device 10 generates an answer using the advanced model. On the other hand, when the user subscribes to the freemium plan, the information providing device 10 generates an answer using the standard model.

[0038] The information providing device 10 provides the generated answer to the user (step S7). The information providing device 10 may accept the prompt and provide the answer via a chat-style user interface. When the answer destination 30 is the vehicle 30-2, the information providing device 10 causes the answer to be displayed on a display (e.g., a navigation device, etc.) mounted on the vehicle 30-2, for example. Alternatively, when the prompt is a request for automatic driving to a destination, the information providing device 10 controls a control device of the vehicle 30-2 to perform automatic driving.

[0039] In addition, if the user subscribes to a premium plan, the information providing device 10 may update the advanced model based on the latest data sources and generate further answers. Such updated advanced models are called real-time models. The information providing device 10 may provide additional answers generated using the real-time models.

[0040] The information providing device 10 continuously acquires data sources and updates the language model during the period from when the information providing device 10 generates an answer to when the information providing device 10 provides the generated answer. By using a real-time model, the information providing device 10 can generate an answer that reflects the difference that occurs during this period.

[0041] Next, when the information providing device 10 provides an answer related to a data source to a user, the information providing device 10 provides a reward to the data source provider 20 of the data source (step S8). For example, the information providing device 10 determines the content of the reward to be provided according to the number of times an answer related to the data source is given (or the number of users). Furthermore, the information providing device 10 may provide different rewards for providing an answer using the standard model and providing an answer using the advanced model. For example, the information providing device 10 provides a higher reward for providing an answer related to a data source via the advanced model than for providing an answer via the standard model.

[0042] In this way, the information providing device 10 collects data sources only from data source providers 20 that satisfy the reliability conditions, thereby avoiding the mixing of low-reliability data sources into the learning of the language model, thereby increasing the reliability of the information provided.

[0043] In the information provision system 1, the information provision device 10 and the data source provider 20 (the server 20-1, the user terminal 20-2, and the vehicle 20-3) incorporate semiconductor chips manufactured by the same manufacturer. Each of these semiconductor chips may perform machine learning or deep learning.

[0044] In addition, the semiconductor chips used in the information providing device 10 and the data source providing source 20 are chips of a size suitable for their respective housings. For example, assuming that the chip size of the information providing device 10 is XL size, the server 20-1 and the vehicle 20-3 are used with L size, and the user terminal 20-2 is used with S or M size. For the data source providing source 20, which has a housing size even smaller than that of the user terminal 20-2, the SS size is used and is put into a SoC (System on a chip) to form a single chip. Note that the number of tiles for each size is 200 tiles for the XL size, 50 tiles for the L size, 20 tiles for the M size, 10 tiles for the S size, and 2 tiles for the SS size, but this is an example and is not limited to the above. If it is desired to attach accessories such as a camera or a microphone to the user terminal 20-2, a chip dedicated to accessories other than the semiconductor chip may be set in an empty space near the semiconductor chip in the SoC.

[0045] In this way, by using semiconductor chips manufactured by the same manufacturer in the information providing device 10 and the data source providing source 20, a closed system can be established between the information providing device 10 and the data source providing source 20, and therefore a secure state can be maintained between the information providing device 10 and the data source providing source 20. In other words, hacking, virus infection, and deep fakes can be avoided with high accuracy between the information providing device 10 and the data source providing source 20, and privacy between the information providing device 10 and the data source providing source 20 can be ensured.

[0046] The configuration of the information providing device 10 will be described with reference to Fig. 2. Fig. 2 is a functional block diagram showing an example of the configuration of the information providing device according to the embodiment.

[0047] As shown in FIG. 2, the information providing device 10 includes a communication unit 11, a storage unit 12, and a control unit 13.

[0048] The communication unit 11 transmits and receives information to and from the data source provider 20 or the answer destination 30 via the network.

[0049] The storage unit 12 is realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk drive (HDD), a solid state drive (SSD), an optical disk, etc. Various programs and various data are stored in the storage unit 12. The storage unit 12 has user information 121, standard model information 122, advanced model information 123, and real-time model information 124.

[0050] The user information 121 is information on the plan to which each user subscribes. For example, the user information 121 is information in which the user ID is associated with a premium plan or a freemium plan.

[0051] Standard model information 122, advanced model information 123, and real-time model information 124 are information such as parameters for constructing a standard model, an advanced model, and a real-time model, respectively. The parameters for constructing a model are, for example, weights and biases of a neural network. In learning a model, these parameters are updated.

[0052] The control unit 13 is a controller, and includes, for example, a microcomputer having a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM, an input / output port, and various other circuits. The control unit 13 may also be configured with hardware such as an integrated circuit, for example, an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). The control unit 13 includes an acquisition unit 31, a determination unit 32, a collection unit 33, a generation unit 34, a provision unit 35, and a supply unit 36.

[0053] The acquisition unit 31 acquires information about the provider from the data source provider 20. The information about the provider is attribute information of the provider (business content, business scale, number of employees, number of registered members) and information about the area where the business is conducted.

[0054] The determination unit 32 determines whether the destination satisfies the reliability condition based on the information on the source. For example, the determination unit 32 determines that the source (newspaper company, news agency, broadcasting company, publisher, net news provider) whose business is news distribution satisfies the reliability condition. In other words, the determination unit 32 determines that the data source of news articles provided by the source whose business is news distribution satisfies the reliability condition because the reliability is high. In addition, the determination unit 32 determines that the source whose business scale is a certain level or more satisfies the reliability condition. The business scale is, for example, the number of news distributions (number of copies in the case of a newspaper company, number of users who have distributed in the case of a news agency, viewer rating in the case of a broadcasting company, number of copies in the case of a publisher, number of distributions or number of registered users in the case of a net news provider). In other words, the determination unit 32 determines that the source whose news articles are distributed to a certain level or more of users satisfies the reliability condition.

[0055] Furthermore, when the data source provider 20 is a device such as a user terminal 20-2 or a vehicle 20-3, the determination unit 32 determines whether or not the manufacturer of the device satisfies a condition regarding reliability. For example, the determination unit 32 determines that a manufacturer with sales amount equal to or greater than a certain amount satisfies the condition regarding reliability.

[0056] The collection unit 33 collects data sources from the data source providers 20 that are determined by the determination unit 32 to satisfy the conditions regarding reliability. The collection unit 33 accumulates the collected data sources in the storage unit 12. The collection unit 33 also continuously collects the data sources asynchronously with the operation of other processing units. That is, the collection unit 33 continues to collect the data sources asynchronously with processing such as training a model and generating and providing an answer.

[0057] The amount of data accumulated in the storage unit 12 by the collection unit 33 changes over time as shown in Fig. 3. Fig. 3 is a diagram for explaining the change in the amount of data.

[0058] The horizontal axis in Fig. 3 is time (hours). The vertical axis in Fig. 3 is the amount of data. The amount of data at the current time t2 is v2. The amount of data at time t1, which is a period T in the past from time t2, is v1. Also, v1 is smaller than v2.

[0059] In addition, the advanced model is trained at time t2, and the time when an answer is generated using the trained advanced model is t2+Δt. Since the collection unit 33 continues to collect data, the amount of data increases by Δv during the period Δt.

[0060] The generation unit 34 generates a language model. Specifically, the generation unit 34 generates a standard model and an advanced model. Specifically, the generation unit 34 learns two language models based on a data source. Specifically, the generation unit 34 learns the standard model based on a past data source. Also, the generation unit 34 learns the advanced model based on a current data source.

[0061] Furthermore, the generation unit 34 updates the parameters of the model through learning. The generation unit 34 may regenerate the model at each learning timing, or may reflect a data source of the difference from the previous learning in the model.

[0062] In addition, the generation unit 34 judges the reliability of the data source collected from the provider, and judges whether or not to use the data source for learning the language model based on the reliability. For example, the generation unit 34 analyzes the contents of the data source (text analysis, voice analysis, image analysis), and if there is a possibility that the contents of the data source are false, prohibits the use of the data source for learning the language model.

[0063] In addition, the generation unit 34 uses the delivery data and comment data included in the data source for model training. This comment data may include supplementary information to the delivery data, more detailed information, information from a different perspective, etc., so by using such comment data for training, it is possible to further improve the accuracy of answers output from the model.

[0064] In addition, the generation unit 34 learns the model based on comments with positive content (positive comments) and comments with negative content (negative comments) in the comment data. Comments with positive content are comments that praise, approve, affirm, or supplement the distributed data. Comments with negative content are comments that criticize, oppose, or deny the distributed data, comments that contain extreme language, and comments with discriminatory content.

[0065] For example, the generation unit 34 prohibits negative comments from being used in model learning among the comment data. This prevents the model from outputting answers based on negative comments, reducing the discomfort felt by users who receive answers.

[0066] Furthermore, when the number of comments in the comment data is equal to or greater than a threshold, the generation unit 34 uses the comment data for model learning. That is, the generation unit 34 determines whether the number of comments in the comment data for each delivery data is equal to or greater than a threshold, and when it is equal to or greater than the threshold, uses the delivery data and the comment data for model learning, and when the number of comments is less than the threshold, uses only the delivery data for learning. This makes it possible to avoid a situation where the diversity of comments cannot be ensured due to a small number of comments.

[0067] In addition, the generation unit 34 adjusts the amount of learning for the model based on the number of positive comments and the number of negative comments in the number of comments. For example, when the ratio of negative comments among all comments in the comment data is equal to or greater than a threshold, the generation unit 34 prohibits the use of all comments or negative comments for learning the model.

[0068] In addition, the generation unit 34 increases the learning amount of distribution data in which the ratio of positive comments is higher than the ratio of negative comments among all comments in the comment data, compared to distribution data in which the ratio of positive comments is lower than the ratio of negative comments. For example, in the case of distribution data in which the ratio of positive comments is lower than the ratio of negative comments, the generation unit 34 prevents content related to the negative comments from being used in learning the model.

[0069] In addition, the generating unit 34 excludes sources that have a predetermined number or more of distribution data with negative comments exceeding a threshold from the collection of data sources. This makes it possible to exclude sources that distribute a lot of distribution data with a lot of negative comments, in other words, sources that distribute unreliable distribution data, from the collection of data sources, thereby increasing the reliability of the collected data sources.

[0070] The providing unit 35 uses the generated language model to generate and provide a response to a prompt received from a user.

[0071] The provision unit 35 determines the plan of the user by referring to the user information 121. The provision unit 35 determines whether the user has subscribed to a premium plan or a freemium plan.

[0072] The providing unit 35 inputs a prompt to a language model and generates an answer. The providing unit 35 generates an answer to the prompt input by the user using one of a plurality of language models having different amounts of data sources used for learning. For example, the providing unit 35 generates an answer using one of a standard model and an advanced model.

[0073] The providing unit 35 provides the answer generated by the language model to the user together with information indicating the provider of the data source used to learn the language model used to generate the answer. For example, the providing unit 35 causes an information providing screen to be displayed on the user terminal. For example, the providing unit 35 may accept the prompt and provide the answer via a chat-style user interface. Furthermore, when the answer providing destination 30 is the vehicle 30-2, the providing unit 35 causes, for example, a display (e.g., a navigation device, etc.) mounted on the vehicle 30-2 to display the answer. Alternatively, when the prompt is a request for autonomous driving to a destination, the providing unit 35 controls a control device of the vehicle 30-2 to perform autonomous driving.

[0074] When an answer related to a data source is provided to a user, the payment unit 36 ​​pays a reward to the data source provider 20 of the data source. For example, the payment unit 36 ​​determines the content of the reward to be paid according to the number of times an answer related to the data source is given (or the number of users). Furthermore, the payment unit 36 ​​may differentiate the reward between the provision via the standard model and the provision via the advanced model. For example, the payment unit 36 ​​makes the reward higher when an answer related to a data source is provided via the advanced model than when the answer is provided via the standard model.

[0075] The flow of the reliability determination process will be described with reference to Fig. 4. Fig. 4 is a flowchart illustrating the flow of the reliability determination process.

[0076] As shown in FIG. 4, the information providing device 10 acquires information on the provider from the data source provider 20 (step S101).

[0077] Next, the information providing device 10 determines whether or not a condition regarding reliability is satisfied based on the information regarding the provider (step S102).

[0078] If the reliability condition is satisfied (step S102: Yes), the information providing device 10 adopts the data source as a provider (step S103) and ends the process. On the other hand, if the reliability condition is not satisfied (step S102: No), the information providing device 10 does not adopt the data source as a provider and ends the process.

[0079] The flow of the pre-training process will be described with reference to Fig. 5. Fig. 5 is a flowchart illustrating the flow of the pre-training process.

[0080] 5, the information providing device 10 continuously acquires data sources (step S201). Next, the information providing device 10 determines whether it is time to generate a model (step S202).

[0081] If it is not the model generation timing (step S202: No), the information providing device 10 returns to step S201 and continues to acquire data sources. On the other hand, if it is the model generation timing (step S202: Yes), the information providing device 10 proceeds to step S203.

[0082] The information providing device 10 trains the standard model using past data sources for a certain period of time (step S203). The information providing device 10 also trains the advanced model using data sources up to the present (step S204). For example, the information providing device 10 trains the advanced model using data sources up to the present, and trains the standard model using data sources up to two years ago.

[0083] The flow of the information provision process will be described with reference to Fig. 6. Fig. 6 is a flowchart illustrating the flow of the information provision process. The information provision device 10 can execute the pre-training process and the information provision process in parallel.

[0084] 6, the information providing device 10 continuously acquires data sources (step S301). The information providing device 10 determines whether or not there is an inquiry from the user (step S302).

[0085] If there is no inquiry from the user (step S302: No), the information providing device 10 returns to step S301 and continues to acquire data sources. On the other hand, if there is an inquiry from the user (step S302: Yes), the information providing device 10 determines the user's plan (step S303).

[0086] If the user's plan is a freemium plan (step S303: freemium), the information providing device 10 proceeds to step S304. On the other hand, if the user's plan is a premium plan (step S303: premium), the information providing device 10 proceeds to step S306.

[0087] In step S304, the information providing device 10 inputs the inquiry content into the standard model and generates an answer. Then, the information providing device 10 provides the generated answer to the user (step S305). Then, the information providing device 10 pays a reward to the provider of the data source related to the answer (step S311), and ends the process.

[0088] In step S306, the information providing device 10 inputs the inquiry content to the advanced model and generates a response, and then provides the generated response to the user (step S307).

[0089] Furthermore, the information providing device 10 updates the advanced model using the latest data source (step S308). The information providing device 10 inputs the inquiry content into the updated advanced model (real-time model) and generates a response (step S309). The information providing device 10 provides the generated response to the user (step S310).

[0090] According to this embodiment, by partnering with companies that are generally recognized as reputable companies and using such companies as providers of data sources, it is possible to increase the reliability of a service that provides answers using a language model.

[0091] Furthermore, according to this embodiment, general answers can be provided to individual free members through a freemium plan, and highly accurate answers can be provided to individual paid members, companies, government and other public institutions through a premium plan.

[0092] 7 is a diagram showing an example of a computer hardware configuration functioning as an information providing device. A program installed in the computer 1200 can cause the computer 1200 to function as one or more "parts" of the device according to the present embodiment, or cause the computer 1200 to execute operations or one or more "parts" associated with the device according to the present embodiment, and / or cause the computer 1200 to execute a process or steps of the process according to the present embodiment. Such a program can be executed by the CPU 1212 to cause the computer 1200 to execute specific operations associated with some or all of the blocks of the flowcharts and block diagrams described herein.

[0093] The computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, and a graphic controller 1216, which are connected to each other by a host controller 1210. The computer 1200 also includes input / output units such as a communication interface 1222, a storage device 1224, a DVD drive, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The DVD drive may be a DVD-ROM drive, a DVD-RAM drive, etc. The storage device 1224 may be a hard disk drive, a solid state drive, etc. The computer 1200 also includes input / output units such as a ROM 1230 and a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0094] The CPU 1212 operates according to a program stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphic controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into itself, and causes the image data to be displayed on the display device 1218.

[0095] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive reads programs or data from a DVD-ROM or the like and provides them to the storage device 1224. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0096] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or a program that depends on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a USB port, a parallel port, a serial port, a keyboard port, a mouse port, and the like.

[0097] The programs are provided by a computer-readable storage medium such as a DVD-ROM or an IC card. The programs are read from the computer-readable storage medium, installed in the storage device 1224, the RAM 1214, or the ROM 1230, which are also examples of computer-readable storage media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and brings about cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be constructed by implementing operations or processing of information according to the use of the computer 1200.

[0098] For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 may execute a communication program loaded in the RAM 1214 and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer area provided in the RAM 1214, the storage device 1224, a DVD-ROM, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer area or the like provided on the recording medium.

[0099] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, a DVD drive (DVD-ROM), an IC card, etc. to be read into the RAM 1214, and perform various types of processing on the data on the RAM 1214. The CPU 1212 may then write back the processed data to the external recording medium.

[0100] Various types of information, such as various types of programs, data, tables, and databases, may be stored in the recording medium and undergo information processing. The CPU 1212 may perform various types of processing on the data read from the RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequence of the program, and write back the results to the RAM 1214. The CPU 1212 may also search for information in a file, database, etc. in the recording medium. For example, when a plurality of entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored in the recording medium, the CPU 1212 may search for an entry whose attribute value of the first attribute matches a specified condition from among the plurality of entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0101] The above-mentioned programs or software modules may be stored in a computer-readable storage medium on the computer 1200 or in the vicinity of the computer 1200. In addition, a recording medium such as a hard disk or a RAM provided in a server system connected to a dedicated communication network or the Internet can be used as a computer-readable storage medium, thereby providing the programs to the computer 1200 via the network.

[0102] The blocks in the flowcharts and block diagrams in the present embodiment may represent stages of a process in which an operation is performed or "parts" of an apparatus responsible for performing the operation. Particular stages and "parts" may be implemented by dedicated circuitry, programmable circuitry provided with computer readable instructions stored on a computer readable storage medium, and / or a processor provided with computer readable instructions stored on a computer readable storage medium. The dedicated circuitry may include digital and / or analog hardware circuits, and may include integrated circuits (ICs) and / or discrete circuits. The programmable circuitry may include reconfigurable hardware circuits, such as, for example, field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like, including AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and memory elements.

[0103] A computer-readable storage medium may include any tangible device capable of storing instructions that are executed by a suitable device, such that a computer-readable storage medium having instructions stored thereon comprises an article of manufacture that includes instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable storage media may include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, and the like. More specific examples of computer-readable storage media may include floppy disks, diskettes, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), electrically erasable programmable read-only memories (EEPROMs), static random access memories (SRAMs), compact disk read-only memories (CD-ROMs), digital versatile disks (DVDs), Blu-ray disks, memory sticks, integrated circuit cards, and the like.

[0104] The computer readable instructions may include either assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0105] Computer readable instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, or to a programmable circuit, either locally or over a local area network (LAN), a wide area network (WAN), such as the Internet, etc., to cause the processor of the general purpose computer, special purpose computer, or other programmable data processing apparatus, or to a programmable circuit, to execute the computer readable instructions to generate means for performing the operations specified in the flowcharts or block diagrams. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0106] Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It is clear to those skilled in the art that various modifications or improvements can be made to the above embodiments. It is clear from the claims that such modifications or improvements can also be included in the technical scope of the present invention.

[0107] It should be noted that the order of execution of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and may be realized in any order unless the output of a previous process is used in a later process. Even if the operational flow in the claims, specifications, and drawings is explained using "first," "next," etc. for convenience, it does not mean that it is essential to perform the process in this order. [Explanation of symbols]

[0108] 10 Information provision device 11 Communications Department 12 Storage section 13 Control section 31 Acquisition Department 32 Judgment section 33 Collection Department 34 Generation part 35 Providing Department 36 Payment Department 121 User Information 122 Standard Model Information 123 Advanced Model Information 124 Real-time model information

Claims

1. A collection unit that collects data sources provided from providers that satisfy reliability conditions; a generation unit that generates a language model by learning the collected data sources, the language model being configured to generate an answer to a prompt input by a user; having The data source is The data includes distribution data distributed to the distribution target users by the provider, and comment data posted by the distribution target users on the distribution data. Information provision device.

2. The generation unit is Among the comment data, comments with negative content are prohibited from being used in learning the language model.

2. The information providing device according to claim 1.

3. The generation unit is When the number of comments in the comment data is equal to or greater than a threshold, the comment data is used for learning the language model.

2. The information providing device according to claim 1.

4. The generation unit is The amount of learning for the language model is adjusted based on the number of positive comments and the number of negative comments in the number of comments in the comment data.

2. The information providing device according to claim 1.

5. The generation unit is If the ratio of negative comments among all comments in the comment data is equal to or greater than a threshold, all of the comments are prohibited from being used for learning the language model.

5. The information providing device according to claim 4.

6. The generation unit is When a ratio of the negative comments among all the comments in the comment data is equal to or greater than a threshold, the negative comments are prohibited from being used for learning the language model.

5. The information providing device according to claim 4.

7. The generation unit is The learning amount of the delivery data in which the ratio of the comments with positive contents is higher than the ratio of the comments with negative contents among all the comments in the comment data is made larger than the learning amount of the delivery data in which the ratio of the comments with positive contents is lower than the ratio of the comments with negative contents.

5. The information providing device according to claim 4.

8. The generation unit is The provider having a predetermined number or more of the distribution data with negative comments exceeding a threshold is excluded from the collection of the data source.

2. The information providing device according to claim 1.

9. An information providing method executed by an information providing device, comprising: A collection step of collecting data sources provided by providers that satisfy reliability conditions; generating a language model by learning the collected data sources, the language model being adapted to generate responses to prompts input by a user; Including, The data source is The data includes distribution data distributed to the distribution target users by the provider, and comment data posted by the distribution target users on the distribution data. Information provision method.

10. A collection step of collecting data sources provided by providers that satisfy reliability criteria; generating a language model by learning the collected data sources, the language model being adapted to generate answers to prompts input by a user; Run the following on your computer: The data source is The data includes distribution data distributed to the distribution target users by the provider, and comment data posted by the distribution target users on the distribution data. Informational program.

Images