Information processing device and information processing method

Abstract

An information processing device 10 according to an embodiment comprises: an acquisition unit 12 that acquires a question from a first user to a second user, background information on a background of the question, and role information on a role of at least one of the first user and the second user; a determination unit 13 that determines, on the basis of the background information and the role information, external information to be referred to; and an output unit 14 that outputs, on the basis of the external information, input information for instructing a generative AI model 31 to generate an answer sentence to the question.

Description

Information processing device and information processing method

[0001] One aspect of the present disclosure relates to an information processing device and an information processing method.

[0002] Patent Document 1 discloses a mechanism for accepting a question from a user and having a trained model output an appropriate answer sentence as an answer to the input question from among multiple answer candidates pre-registered in a database.

[0003] JP 2024-6940 A

[0004] According to the mechanism described in Patent Document 1, it is possible to provide a user with an appropriate answer to a user's question, but in order to improve user convenience, it is necessary to sufficiently improve the accuracy of the answer. Furthermore, there is a need to apply the mechanism described above to a situation in which one user asks another user a question during a specific conversation between multiple users, so as to automatically and accurately generate a reply that should be made from the other user to the first user in order to facilitate communication between the users.

[0005] Therefore, an object of one aspect of the present disclosure is to provide an information processing device and an information processing method that can automatically and accurately provide an appropriate answer to a question from one user to another user.

[0006] An information processing device according to one aspect of the present disclosure includes an acquisition unit that acquires a question from a first user to a second user, background information related to the background of the question, and role information related to the role of at least one of the first user and the second user, a determination unit that determines external information to refer to based on the background information and the role information, and an output unit that outputs input information to instruct a generative AI model to generate an answer sentence to the question based on the external information.

[0007] An information processing method according to another aspect of the present disclosure includes the steps of acquiring a question from a first user to a second user, background information relating to the background of the question, and role information relating to the role of at least one of the first user and the second user, determining external information to refer to based on the background information and the role information, and outputting input information to instruct a generative AI model to generate an answer sentence to the question based on the external information.

[0008] According to one aspect of the present disclosure, it is possible to automatically and accurately provide an appropriate answer to a question posed by one user to another user.

[0009] 4 is a block diagram showing an overall configuration of a system including an information processing device according to an embodiment; FIG. 5 is a diagram showing an example of user information; FIG. 6 is a diagram showing an example of a prompt used in a mode information determination process; FIG. 7 is a diagram showing an example of a prompt for instructing the generation of an answer sentence; FIG. 8 is a diagram showing an example of an answer sentence generated by the prompt of FIG. 4; FIG. 9 is a diagram showing an example of an additional prompt; FIG. 10 is a flowchart showing an example of the operation of an information processing device; and FIG. 11 is a diagram showing an example of a hardware configuration of an information processing device.

[0010] Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0011] Fig. 1 is a diagram showing the overall configuration of a system including an information processing device 10 according to one embodiment. The system shown in Fig. 1 includes the information processing device 10, a user terminal 20, and a generation AI server 30. The information processing device 10 is configured to be able to communicate with each of the user terminal 20 and the generation AI server 30 via any communication network. The information processing device 10 functions as a so-called RAG (Retrieval-Augmented Generation) system.

[0012] The information processing device 10 executes a process for causing the generative AI model to generate an answer sentence to a question from a speaker user (first user) to a listener user (second user). The information processing device 10 is also configured to generate and output input information for instructing the generative AI server 30 based on information received from the user terminal 20 (in this embodiment, background information and user information, which will be described later).

[0013] As an example, the input information is information that constitutes at least a part of the prompt input to the generation AI model 31. The generation AI server 30 receives a prompt including such input information and inputs the prompt to the generation AI model 31, thereby generating and outputting an answer sentence to a question from a speaker user to a listener user. The answer sentence output by the generation AI model 31 may be transmitted directly to the user terminal 20, or may be transmitted to the user terminal 20 via the information processing device 10. Alternatively, the answer sentence may be accumulated (stored) on any server that the user can access via the user terminal 20, etc. In this case, the user can view the answer sentence generated by the generation AI model 31 by accessing the server at any time.

[0014] The user terminal 20 is, for example, a terminal device used by either a speaker user or a hearing user. While the method of using the services provided by the information processing device 10 is not particularly limited, the following two methods are typically considered. The first method is a case where the speaker user is the user. For example, the first method includes a case where the speaker user obtains a response from a virtual hearing user (the generation AI model 31) when the hearing user is absent (when there is no response from the hearing user). The second method is a case where the hearing user is the user. For example, the second method includes a case where the hearing user causes the generation AI model 31 to output a response to a question received from the speaker user, and the hearing user confirms the response (and modifies it as appropriate) and responds to the speaker user. In the case of the first method, the user terminal 20 is a terminal device used by the speaker user. In the case of the second method, the user terminal 20 is a terminal device used by the hearing user.

[0015] The user terminal 20 may take any form, such as a personal computer, a smartphone, a tablet terminal, a feature phone, a server device, a game console, etc. For example, if the service provided from the information processing device 10 to the user terminal 20 is in the form of a web service, the user terminal 20 may be a client terminal equipped with a web browsing function. Note that while only one user terminal 20 is illustrated in FIG. 1 , the information processing device 10 may be configured to be accessible from multiple user terminals 20 used by multiple users. In other words, the system illustrated in FIG. 1 may include multiple user terminals 20.

[0016] The generation AI server 30 includes a generation AI model 31, and is a device that generates and provides content using the generation AI model 31. The generation AI model 31 is a model that can generate content (in this embodiment, information indicating the content of a response sentence to a speaker user) in response to input of a prompt (input information) output by the information processing device 10, according to any one or a combination of the instructions, context, question, and output format indicated by the prompt, and return the content as response information.

[0017] A prompt is information indicating an instruction or question input to the generative AI model 31 in an interactive system such as a dialogue with the generative AI model 31 or a command line interface (CLI). The prompt can include various types of input information. In this case, the generative AI model 31 generates response information targeted at the various types of input information. The type and format of the input information are not particularly limited, and the input information may include, for example, data files with file names including a predetermined extension, such as text data, image data, application-related data, audio data, video data, and still image data. Application-related data is data such as document data, table data, and graph data that can be processed by a default application program.

[0018] The generative AI model 31 may be configured to include, for example, a large-scale language model (LLM) and a user interface (UI) for interacting with a user. The generative AI model 31 may be an interactive AI model capable of chatting with a user via text or voice. Examples of such generative AI models 31 include ChatGPT, GPT-3.5, GPT-4V, PaLM2, and Gemini. However, the generative AI model 31 may be an AI model other than the large-scale language model described above.

[0019] The generative AI model 31 may be placed in a generative AI server 30 separate from the information processing device 10, as shown in FIG. 1, or may be placed in the information processing device 10. Alternatively, the generative AI model 31 may be placed in a user terminal 20, for example, like tsuzumi. In this case, each function of the information processing device 10 may also be placed in the user terminal 20. That is, the user terminal 20 itself may function as the information processing device 10 of this embodiment.

[0020] The information processing device 10 includes a storage unit 11, an acquisition unit 12, a determination unit 13, and an output unit 14. In this embodiment, an example will be described in which the speaker user is the president of a company and the hearing user is the speaker user's secretary. That is, an example will be described in which the speaker user's role is "president" and the hearing user's role is "secretary." However, the combination of speaker users and hearing users is not limited to the above. That is, the combination of the roles of speaker users and hearing users may be a combination other than the above.

[0021] The storage unit 11 stores various data handled by the information processing device 10. The storage unit 11 includes a knowledge database that is a search destination for information used to generate prompts for the generation AI model 31. In this embodiment, the storage unit 11 that functions as such a knowledge database is implemented on the information processing device 10, but the knowledge database may also be implemented outside the information processing device 10 (for example, on a network). In this embodiment, the storage unit 11 has a user information DB and an internal company information DB.

[0022] The user information DB stores user information about users who belong to the companies of the speaker and listener users. FIG. 2 is a diagram showing an example of user information. As shown in FIG. 2, in this embodiment, the user information is information that associates, for each user, identification information (user ID) for identifying the user, the role of the user, and the access rights of the user. The top record (user information for one person) in the example of FIG. 2 indicates that the role of the user with user ID "U001" is "president" and includes information on the access rights of the user.

[0023] "Role" is information indicating the role, position, etc. of a user. In this embodiment, examples of roles include "president" and "secretary." As another example, if the speaker user is a student and the hearing user is a teacher, roles such as "student" and "teacher" may be set.

[0024] "Access rights" is information indicating the data, range, etc. that a user can access. In this embodiment, the access rights information is used to prevent the generation of an answer to a question from being based on information that the user (speaker user or listener user) would not normally be able to access (for example, preventing the answer from including information that the user would not normally be able to know) when the generation AI model 31 is made to generate an answer to a question.

[0025] The access rights may be in one-to-one correspondence with the roles. That is, the contents of the access rights may be set for each role. In this case, the user information may be composed of first table information in which "user IDs" and "roles" are associated with each other, and second table information in which "roles" and "access rights" are associated with each other.

[0026] The internal information DB stores external information to be referenced when the generation AI model 31 generates a response sentence. The external information may include, for example, various information managed within the company. Examples of external information include response information, which is a record of past interactions with customers, email information, scheduler information, and minutes of past meetings. Each of multiple items (e.g., information about individual schedules) included in the external information (e.g., scheduler information) is associated with information regarding importance. The importance may be set by a user (e.g., a speaker user or a listener user) or may be automatically set based on predetermined rules according to the user's role. For example, the importance of each meeting schedule may be automatically set according to the degree to which participation of a role (e.g., the president) predetermined for each type of meeting is required. For example, if participation is required, the importance is set higher than if participation is not required (e.g., if participation is preferred). In this embodiment, the in-house information DB also stores in advance information (prompt template information) that serves as the basis for prompts used in the mode information determination process and the external information determination process, which will be described later.

[0027] The acquisition unit 12 acquires a question from a speaker user to a hearer user, background information relating to the background of the question, and role information relating to the role of at least one of the speaker user and the hearer user. For example, the acquisition unit 12 acquires the question and background information from the user terminal 20, as well as the user IDs of the speaker user and the hearer user.

[0028] The question information (question data) acquired by the acquisition unit 12 may be text data or voice data. In the latter case, when text-format question data is required for input to the later-described generative AI model 31, the acquisition unit 12 may convert the question data into text format by performing a known voice recognition process or the like (or by having an external server or the like perform a voice recognition process or the like).

[0029] Furthermore, the acquisition unit 12 acquires the roles and access rights of each of the speaker user and the listener user by referring to the storage unit 11 (user information DB) using the user IDs of each of the speaker user and the listener user as a search key. Thus, in this embodiment, the acquisition unit 12 acquires role information related to the roles of both the speaker user and the listener user (i.e., both the roles of the speaker user and the roles of the listener user). Furthermore, the acquisition unit 12 acquires authority information related to the access rights of both the speaker user and the listener user (i.e., both the access rights of the speaker user and the access rights of the listener user).

[0030] The question includes, for example, a demonstrative pronoun. Examples of such questions include questions that include a demonstrative pronoun (in this example, "that") in the sentence, such as "When was the next thing?" or "By the way, what happened to that?"

[0031] The background information is information relating to the background (background circumstances), situation (scene), etc. of when a question was asked. For example, the background information may be information indicating how the question was asked. As an example, the background information is information relating to the content of a conversation between a speaker user and a hearer user before the speaker user asked the hearer user a question. In other words, the background information may be information indicating the context that led to the speaker user asking the hearer user a question.

[0032] Below are examples of background information showing the content of a conversation between a speaker user (the president) and a listener user (the secretary), and examples of questions. Note that in the following [Examples of Background Information], (Speaker) indicates the content of what the speaker user said, and (Listener) indicates the content of what the listener user said.

[0033] [Example of background information] (Speaker) Mr. M from the KK department gave a long explanation of the measures. (Listener) Thank you for your hard work. (Speaker) However, if we can't see the results of the measures taken so far in the KK department, we'll be hesitant to take the next measures. (Listener) It's unfortunate that there hasn't been much progress in terms of information. [Example of question] (Speaker) When was the next one?

[0034] The determination unit 13 determines the external information to be referenced based on the background information and role information. In this embodiment, the determination unit 13 executes a process of determining mode information indicating the source of the external information to be referenced based on the background information and role information (hereinafter referred to as a "mode information determination process"). Next, the determination unit 13 executes a process of determining the external information from among the information included in the source indicated by the mode information (hereinafter referred to as an "external information determination process"). An example of the mode information determination process and the external information determination process will be described below.

[0035] [Example of Mode Information Determination Process] The determination unit 13 may determine mode information, for example, by using the generative AI model 31. For example, the determination unit 13 may input a prompt (instruction information) including at least background information, role information, and content instructing the generative AI model 31 to output mode information corresponding to the background information and role information, to the generative AI model 31, and determine the mode information based on the output result from the generative AI model 31.

[0036] 3 shows an example of a prompt P1 generated by the determination unit 13 in the mode information determination process, and an example of response information A1 output from the generation AI model 31 based on the prompt P1. In this example, the prompt P1 is composed of five parts: "#Previous interactions," "#Role information," "#Current time," "#Mode candidates," and "#Inquiry."

[0037] "#Previous interactions" corresponds to the background information and question acquired by the acquisition unit 12. The part up to "You: It's a shame there hasn't been much progress in terms of information." corresponds to the background information, and "Me: When was that next thing?" corresponds to the question. "#Role information" corresponds to the role information acquired by the acquisition unit 12. In this example, information indicating the roles of both the speaker user and the listener user is included in the prompt P1 (#Role information). Since it is assumed that the generation AI model 31 will answer the question, "#Current time" is, for example, the time indicating the time when the question was asked. The information indicating the current time may be transmitted from the user terminal 20 to the acquisition unit 12. Alternatively, the time corresponding to the time when the acquisition unit 12 received the question, background information, etc. from the user terminal 20 may be set as the current time.

[0038] "#mode candidate" is information indicating candidates for mode information. The mode candidates may be, for example, predetermined. In this example, "reference to customer reception information" indicates a mode for referring to "customer reception information" managed as in-house information, "reference to scheduler" indicates a mode for referring to the speaker user's scheduler (individual registered schedule information), "reference to email" indicates a mode for referring to the speaker user's email, and "reference to minutes" indicates a mode for referring to the minutes of a meeting attended by the speaker user. These are pre-registered as mode candidates.

[0039] "#Inquiry" is information corresponding to the main instruction content to the generative AI model 31. In this example, the instruction content is set as follows: "Please refer to each piece of information above and select from the 'mode candidates' what topic the conversation is about." In other words, the instruction content is an inquiry as to which mode (topic) listed in "#Mode candidates" the conversation in "#Previous Interactions" is likely to be related to, based on information such as "#Previous Interactions," "#Role Information," and "#Current Time."

[0040] 3, a prompt P1 including the above-described information is input to the generating AI model 31, which in turn outputs mode information indicating scheduler reference (i.e., mode information indicating the scheduler as the source of external information) as response information A1. The determining unit 13 can determine the mode information (scheduler reference in this example) based on the response information A1 output from the generating AI model 31 in this manner.

[0041] In the above prompt P1, the utterance content of the speaker user is written as "I" utterance, and the utterance content of the hearing user is written as "you" utterance. This is for the following reason. If the party giving instructions to the generation AI model 31 (the information processing device 10) is considered to be "I," then the generation AI model 31 corresponds to "you." Furthermore, since the answer sentence that the generation AI model 31 ultimately generates is positioned as the answer content from the hearing user, it is preferable to make the role of the generation AI model 31 correspond to the hearing user. For this reason, in order to correspond the hearing user to the generation AI model 31, in prompt P1, the speaker user is associated with "I," while the hearing user is associated with "you."

[0042] [Example of External Information Determination Process] The determination unit 13 determines the external information to be referenced (i.e., the external information to be included in prompt P2, which will be described later) from the information included in the source indicated by the mode information determined in the mode information determination process. The determination unit 13 may also determine the external information based on the mode information and the authority information. In the example of FIG. 3 , the mode information is "scheduler reference," the role of the speaker user is "president," and the access rights (authority information) of the speaker user are the authority associated with "president." In this case, the determination unit 13 may determine, as the external information to be referenced, information included in the scheduler information of the speaker user (president) (information that can be referenced with the speaker user's access rights). Through this external information determination process, for example, the information in portion P21 in FIG. 4 is determined as the external information to be referenced.

[0043] In this example, scheduler information containing multiple schedules (for example, six) is acquired. Each item (record indicating an individual schedule) included in the scheduler information is expressed in a data format with five elements separated by commas, such as "data type, scheduled period, participants, scheduled content, importance." In this example, the data type is "scheduler," indicating that it is scheduler information.

[0044] The output unit 14 outputs input information for instructing the generative AI model 31 to generate an answer sentence to the question based on the external information determined by the determination unit 13. In the present embodiment, as an example, the output unit 14 outputs (generates) the entire prompt to be input to the generative AI model 31 as input information. The generative AI model 31 generates and outputs an answer sentence by inputting the prompt output from the output unit 14. The output unit 14 then transmits the answer sentence received from the generative AI model 31 to the user terminal 20. The data indicating the answer sentence output by the output unit 14 may be text data or voice data synthesized based on the answer sentence. For example, the information processing device 10 may be configured as an interactive robot that outputs an answer sentence as voice data.

[0045] FIG. 4 shows an example of a prompt P2 generated by the output unit 14. Note that the prompt P2 is an example of a prompt that assumes that the generation AI model 31 continues to store (retain) the content of the previously input prompt P1. For this reason, a portion P23 of the prompt P2 quotes the "previous exchange" included in the prompt P1. For example, if the generation AI model to which the prompt P2 for generating a response sentence is input is different from the generation AI model to which the prompt P1 is input in the mode information determination process described above, the prompt P2 may include the matter included in the prompt P1. In this embodiment, it is assumed that the generation AI model to which the prompt P1 is input and the generation AI model to which the prompt P2 is input are the same, and therefore the prompt P2 is positioned as a further instruction following the instruction by the prompt P1.

[0046] 4, prompt P2 is composed of four parts: "#mode," "#scheduler information," "#schedule adjustment matters to be observed," and "#inquiry." "#mode" is information indicating the mode information (in this example, scheduler reference) determined in the mode information determination process of the determination unit 13 described above.

[0047] The "# scheduler information" shown in part P21 is external information (information acquired from the in-house information DB of the storage unit 11) determined in the external information determination process of the determination unit 13 described above.

[0048] The "#Strict Observations Regarding Schedule Adjustment" shown in portion P22 is an example of a reply generation rule predetermined according to mode information. That is, the output unit 14 is configured to include, in the prompt P2, a reply generation rule corresponding to the mode information determined by the determination unit 13. As a result, in the example of FIG. 4, a reply generation rule (portion P22) corresponding to "Refer to the scheduler" is included.

[0049] As an example, part P22 includes the following two items, which are predetermined answer generation rules for mode information (see scheduler): If there is a conflict in the schedule, select the one with the highest importance. For a conflicting appointment that could not be selected, refer to a similar appointment that will be held later.

[0050] The "#inquiry" shown in part P23 is information corresponding to the instruction content to the generation AI model 31. In this example, the instruction content is set as follows: "Regarding the theme described in the 'mode', please refer to the 'scheduler information' and be sure to fulfill the 'scheduling adjustment requirements', and then generate an answer to the last question from 'me', including the content indicated by 'that' included in the 'previous interactions'."

[0051] As in the example of Fig. 4, the output unit 14 may output a prompt P2 including instructions to generate an answer sentence based on predetermined answer sentence generation rules (part P22) according to the mode information. In the example of Fig. 4, the instruction included in part P23, "Please be sure to satisfy the 'Strict Observations Regarding Schedule Adjustment'...and then generate an answer sentence to the last question from 'I'," corresponds to the above instruction.

[0052] The answer generation rule (part P22) may also include instructions to consider the importance of each of multiple items of external information (scheduler information in part P21). In the example of Fig. 4, the instruction "If there is a schedule conflict, select the one with the highest importance" included in part P22 corresponds to the instruction.

[0053] The output unit 14 may also output a prompt P2 including a content instructing that the answer sentence include the estimated result of the information indicated by the demonstrative pronoun included in the question. In the example of FIG. 4, the content of the instruction corresponds to the above-mentioned content, which is included in the part P23, "including the content indicated by 'that' included in 'the previous exchanges'."

[0054] 5 shows an example of an answer sentence A2 generated by the generative AI model 31 in response to input of prompt P2 in FIG. 4. In this example, answer sentence A2 includes the estimation result that "that" refers to the next "regular executive meeting." Furthermore, by taking into consideration the importance of each item in the external information (scheduler information in part P21) and the answer sentence generation rule in part P22, answer sentence A2 includes content that refers to a personnel consultation that conflicts with the next regular executive meeting. Furthermore, since the personnel consultation is considered to have priority, answer sentence A2 includes the possibility that the speaker user's attendance at the next regular executive meeting will be "from 1:00 PM to 2:30 PM on June 12, 2024."

[0055] The output unit 14 may output an additional prompt based on the content of the answer sentence A2 obtained as described above in order to output more concise answer information to the user. The additional prompt is instruction information for instructing the generation AI model 31 to generate answer information to be output to the user.

[0056] FIG. 6 is a diagram showing an example of the additional prompt P3 mentioned above. In this example, prompt P3 is composed of three parts: "#Interaction so far," "#Information," and "#Process." "#Interaction so far" is the same as the information contained in prompt P1. "#Information" stores the content of answer sentence A2. "#Process" contains the instruction "Summarize the 'Information' and generate a concise response to 'Your' 'Interaction so far." Such instruction may be embedded in advance as a template.

[0057] The output unit 14 inputs the prompt P3 to the generative AI model 31, thereby acquiring the answer information output by the generative AI model 31. For example, by inputting the prompt P3 to the generative AI model 31, answer information indicating the content, "The next regular executive meeting is scheduled to start at 1:00 PM on June 5, 2024, but because it overlaps with a personnel consultation, it will probably start at 1:00 PM on June 12," is output by the generative AI model 31. The output unit 14 may transmit answer information in which the content of the answer sentence A2 is simplified in this way to the user terminal 20.

[0058] Next, an example of the operation (information processing method) of the information processing device 10 will be described with reference to Fig. 7. Fig. 7 is a flowchart showing an example of the operation of the information processing device 10.

[0059] In step S1, the acquisition unit 12 acquires a question from a speaker user to a hearer user, background information (in this embodiment, the content of the conversation between the speaker user and the hearer user immediately before the question was asked), and role information of the speaker user and the hearer user. In this embodiment, the acquisition unit 12 also acquires the user IDs of the speaker user and the hearer user.

[0060] In step S2, the acquisition unit 12 acquires authority information regarding the access rights of the speaker user and the hearer user. As an example, the acquisition unit 12 can acquire the authority information by referring to the storage unit 11 (user information DB) using the user IDs of the speaker user and the hearer user acquired in step S1 as search keys.

[0061] In step S3, the determination unit 13 determines mode information based on the background information and role information. In this embodiment, the determination unit 13 determines mode information by executing the above-mentioned "mode information determination process." As an example, as shown in FIG. 3, "scheduler reference" is determined as the mode information.

[0062] In step S4, the determination unit 13 determines the external information based on the mode information and authority information determined in step S3. In this embodiment, the determination unit 13 determines the external information by executing the above-mentioned "external information determination process." As an example, the information shown in part P21 in FIG. 4 is determined and acquired as the external information.

[0063] In step S5, the output unit 14 generates and outputs a prompt P2 (see FIG. 4) based on the external information determined in step S4. That is, the output unit 14 generates the prompt P2 and outputs the prompt P2 as input information for the generation AI server 30 (generation AI model 31).

[0064] In step S6, the output unit 14 acquires the answer sentence A2 (see FIG. 5) returned from the generative AI model 31 and transmits the answer sentence A2 to the user terminal 20. Alternatively, the output unit 14 may generate an additional prompt P3 (see FIG. 6) based on the answer sentence A2, input the prompt P3 into the generative AI model 31, acquire the output answer information, and transmit the answer information to the user terminal 20. The answer sentence A2 may be transmitted to the user terminal 20 as is, but by transmitting answer information that has been modified to be more concise as described above to the user terminal 20, a smoother and more natural interaction (conversation) with the user can be realized.

[0065] In the information processing device 10 and information processing method described above, background information regarding the background circumstances of the question and role information regarding the roles of at least one of the speaker user and the listener user (in this embodiment, both) are acquired along with a question from a speaker user to a listener user. Based on this background information and role information, external information to be referenced in order to accurately generate a reply to the question can be appropriately acquired. Then, by instructing the generation AI model 31 to generate a reply based on the external information, an appropriate reply to the question from the speaker user to the listener user can be automatically and accurately generated. By providing the reply to the user (speaker user or listener user), convenience for the speaker user or listener user can be improved and communication between the speaker user and listener user can be facilitated.

[0066] In this embodiment, the acquisition unit 12 acquires a question including a demonstrative pronoun. The output unit 14 outputs input information (prompt P2) including content instructing that an estimation result of information indicated by the demonstrative pronoun be included in an answer sentence. With the above configuration, when a demonstrative pronoun such as "that" is included in a question, a highly accurate answer sentence based on the estimation result of the content indicated by the demonstrative pronoun can be obtained.

[0067] In this embodiment, the background information is information about the content of the conversation between the speaker user and the listener user before the speaker user asks the question. With the above configuration, the generative AI model 31 can accurately understand the background circumstances (context) of the question based on the content of the conversation immediately before the question. As a result, the accuracy of the answer sentence output from the generative AI model 31 can be effectively improved.

[0068] In this embodiment, the role information includes information about the roles of both the speaker user and the listener user. According to the above configuration, the generative AI model 31 can understand the relationship between the speaker user and the listener user based on the role information, thereby allowing the generative AI model 31 to generate a more appropriate reply from the listener user to the speaker user. For example, it is possible to appropriately adjust not only the content of the information included in the reply but also formal aspects such as wording based on the relationship between the speaker user and the listener user. Note that the role information may be information indicating only the role of at least one of the speaker user and the listener user (e.g., the role of the speaker user "President"). Even when only the role of one of the speaker user and the listener user is known, providing the generative AI model 31 with information about the role of at least one of the speaker user and the listener user can be expected to improve the accuracy and naturalness of the content of the reply according to the role.

[0069] In this embodiment, the determination unit 13 determines mode information indicating the source of the external information to be referenced based on the background information and role information, and determines the external information from the information included in the source indicated by the mode information. According to the above configuration, the source of the external information to be referenced (mode information) can be accurately determined based on information about the background of the question (background information) and role information of at least one of the speaker user and the listener user (in this embodiment, both). As a result, the quality of the external information used to generate the answer sentence (including in the prompt P2) can be improved, and the quality of the answer sentence generated by the generation AI model 31 can be appropriately improved.

[0070] In this embodiment, the acquisition unit 12 acquires authority information regarding the access authority of at least one of the speaker user and the hearing user (in this embodiment, both). The determination unit 13 determines the external information based on the mode information and the authority information. With the above configuration, for example, it is possible to determine, as external information, only data to which the speaker user or the hearing user has access rights. In this embodiment, both the speaker user who is the president and the hearing user who is his secretary have access rights to the president's scheduler information. Therefore, the determination unit 13 determines (acquires) the president's scheduler information (part P21 of FIG. 4 ) accessible to the speaker user and the hearing user as external information. If the hearing user's role is not the president's secretary but a role (e.g., "general employee") that does not have access to the president's scheduler information, it would be inappropriate for the hearing user's response to include the president's scheduler information. According to the above configuration, the determination unit 13 determines external information based on authority information (for example, by determining as external information information to which both the speaker user and the listener user have access), making it possible for the generation AI model 31 to generate an answer sentence so that inappropriate situations such as those described above do not occur.

[0071] In this embodiment, the determination unit 13 inputs instruction information (prompt P1) including at least background information, role information, and content instructing the generative AI model 31 to output mode information corresponding to the background information and role information, and determines mode information based on the output result from the generative AI model 31. In this embodiment, the determination unit 13 executes the above-mentioned "mode information determination process." With the above configuration, mode information can be determined easily and accurately. More specifically, there is no need to define rule-based processing in advance, and the generative AI model 31 can be caused to appropriately determine mode information.

[0072] In this embodiment, the output unit 14 outputs input information (prompt P2) including instructions (instructions included in part P23) to generate an answer sentence based on predetermined answer sentence generation rules (part P22) according to the mode information. The above configuration enables the generative AI model 31 to generate more appropriate and useful answer sentences based on predetermined rules according to the determined mode information.

[0073] Furthermore, in this embodiment, the external information (scheduler information) includes multiple items (individual schedules) each associated with information regarding importance, and the answer generation rule includes content instructing the generation AI model 31 to take into consideration the importance of each of the multiple items. According to the above configuration, when the external information to be referenced includes multiple items, the generation AI model 31 can be instructed to reference each item according to its importance, thereby enabling the generation AI model 31 to generate more appropriate and useful answers.

[0074] For example, the acquisition unit 12 acquires information about the roles and access rights of each user based on the user IDs of the speaker user and the hearer user, but the acquisition unit 12 may acquire information indicating the roles or access rights of the speaker user or the hearer user directly from the user terminal 20.

[0075] Furthermore, the generation AI model 31 to which prompt P1 is input and the generation AI model 31 to which prompt P2 is input may be the same or different. The generation AI model 31 to be used may be set in advance in the information processing device 10, or may be set based on a specification from the user of the user terminal 20. In the above embodiment, the template information on which prompts P1 and P2 are based is stored in the memory unit 11 (internal information DB), but may be provided to the information processing device 10 by the user of the user terminal 20.

[0076] Furthermore, additional processing may be added as appropriate to the processing of the information processing device 10 described above. For example, in the mode information determination processing, if the information (e.g., background information) initially acquired from the user terminal 20 is insufficient and mode information cannot be determined with sufficient accuracy, the determination unit 13 may transmit instruction information instructing the user (user terminal 20) to add the missing information. Alternatively, the determination unit 13 may include instruction content such as "If information for determining mode information is insufficient, please generate a question to request additional information from the user" in the prompt P1, thereby causing the generative AI model 31 to output the question as needed.

[0077] [Note] The present disclosure includes the following aspects.

[0078] [1] An information processing device comprising: an acquisition unit that acquires a question from a first user to a second user, background information related to the background of the question, and role information related to the role of at least one of the first user and the second user; a determination unit that determines external information to refer to based on the background information and the role information; and an output unit that outputs input information to instruct a generative AI model to generate an answer sentence to the question based on the external information.

[0079] [2] The information processing device of [1], wherein the acquisition unit acquires the question including a demonstrative pronoun, and the output unit outputs the input information including content instructing that the estimated result of the information indicated by the demonstrative pronoun be included in the answer sentence.

[0080] [3] An information processing device according to [1] or [2], wherein the background information is information about the content of the conversation between the first user and the second user before the first user asked the question.

[0081] [4] The information processing device of any one of [1] to [3], wherein the role information includes information about the roles of both the first user and the second user.

[0082] [5] An information processing device according to any one of [1] to [4], wherein the determination unit determines mode information indicating the source of external information to be referenced based on the background information and the role information, and determines the external information from among the information included in the source indicated by the mode information.

[0083] [6] The information processing device of [5], wherein the acquisition unit further acquires authority information regarding the access authority of at least one of the first user and the second user, and the determination unit determines the external information based on the mode information and the authority information.

[0084] [7] An information processing device according to [5] or [6], wherein the determination unit inputs instruction information including at least the background information, the role information, and content instructing the output of the mode information according to the background information and the role information into a generative AI model, and determines the mode information based on the output result from the generative AI model.

[0085] [8] An information processing device according to any one of [5] to [7], wherein the output unit outputs the input information including content instructing the generation of the answer sentence based on predetermined answer sentence generation rules according to the mode information.

[0086] [9] The information processing device of [8], wherein the external information includes a plurality of items each associated with information regarding importance, and the answer generation rule includes content that instructs taking into consideration the importance of each of the plurality of items.

[0087]

[10] An information processing method including the steps of: acquiring a question from a first user to a second user, background information relating to the background of the question, and role information relating to the role of at least one of the first user and the second user; determining external information to refer to based on the background information and the role information; and outputting input information for instructing a generative AI model to generate an answer sentence to the question based on the external information.

[0088] [Definition of Terms, etc.] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and software. Furthermore, the method of realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are directly or indirectly connected (e.g., using wires, wirelessly, etc.) and these multiple devices. The functional block may also be realized by combining software with the single device or the multiple devices.

[0089] Functions include, but are not limited to, judging, determining, calculating, computing, processing, deriving, investigating, searching, verifying, receiving, transmitting, outputting, accessing, resolving, selecting, choosing, establishing, comparing, expecting, anticipating, regarding, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating, mapping, and assigning.

[0090] For example, the information processing device 10 according to an embodiment of the present disclosure may function as a computer that performs the information processing method of the present disclosure. Fig. 8 is a diagram illustrating an example of the hardware configuration of the information processing device 10 according to an embodiment of the present disclosure. The information processing device 10 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

[0091] In the following description, the term "apparatus" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the information processing device 10 may be configured to include one or more of the apparatuses shown in FIG. 8, or may be configured to exclude some of the apparatuses.

[0092] Each function of the information processing device 10 is realized by loading specified software (programs) onto hardware such as the processor 1001 and memory 1002, causing the processor 1001 to perform calculations, control communication via the communication device 1004, and control at least one of reading and writing data in the memory 1002 and storage 1003.

[0093] The processor 1001 controls the entire computer by running, for example, an operating system, and may be configured as a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc.

[0094] The processor 1001 also reads programs (program code), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes in accordance with the programs. The programs used are those that cause a computer to execute at least some of the operations described in the above-described embodiments. For example, each functional unit of the information processing device 10 (e.g., the determination unit 13, etc.) may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and similar implementations may be used for other functional blocks. While the above-described various processes have been described as being executed by one processor 1001, they may also be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The programs may also be transmitted from a network via a telecommunications line.

[0095] The memory 1002 is a computer-readable recording medium and may be configured, for example, by at least one of a read-only memory (ROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a random access memory (RAM), etc. The memory 1002 may also be called a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store executable programs (program codes), software modules, etc. for implementing an information processing method according to an embodiment of the present disclosure.

[0096] Storage 1003 is a computer-readable recording medium, and may be composed of at least one of, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray (registered trademark) disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy (registered trademark) disk, a magnetic strip, etc. Storage 1003 may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other appropriate medium including at least one of memory 1002 and storage 1003.

[0097] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also called, for example, a network device, a network controller, a network card, or a communication module.

[0098] The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that outputs to the outside. The input device 1005 and the output device 1006 may be integrated into one device (e.g., a touch panel).

[0099] Furthermore, each device, such as the processor 1001 and the memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or may be configured using different buses between each device.

[0100] The information processing device 10 may also be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 1001 may be implemented using at least one of these pieces of hardware.

[0101] Although the present embodiment has been described in detail above, it is clear to those skilled in the art that the present embodiment is not limited to the embodiment described in this specification. The present embodiment can be implemented in modified and altered forms without departing from the spirit and scope of the present invention as defined by the claims. Therefore, the description in this specification is intended to be illustrative and does not have any limiting meaning on the present embodiment.

[0102] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless it is consistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the particular order presented.

[0103] Input and output information may be stored in a specific location (for example, memory) or may be managed using a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be transmitted to another device.

[0104] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).

[0105] The aspects / embodiments described in this disclosure may be used alone, in combination, or switched depending on the implementation. Notification of predetermined information (e.g., notification that "X is true") is not limited to explicit notification, but may be implicit (e.g., not notifying the predetermined information).

[0106] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0107] Software, instructions, information, etc. may also be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then these wired and / or wireless technologies are included within the definition of transmission media.

[0108] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0109] Furthermore, the information, parameters, etc. described in this disclosure may be expressed using absolute values, may be expressed using relative values ​​from a predetermined value, or may be expressed using other corresponding information.

[0110] The names used for the parameters described above are not intended to be limiting in any way. Furthermore, the mathematical formulas and the like that use these parameters may differ from those explicitly disclosed in this disclosure. The various information elements may be identified by any suitable names, and the various names assigned to these various information elements are not intended to be limiting in any way.

[0111] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0112] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed or that the first element must in some way precede the second element.

[0113] When the terms "include," "including," and variations thereof are used in this disclosure, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, when the term "or" is used in this disclosure, it is not intended to be an exclusive or.

[0114] In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.

[0115] In the present disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "coupled" may also be interpreted in the same way as "different."

[0116] 10...information processing device, 11...memory unit, 12...acquisition unit, 13...determination unit, 14...output unit, 20...user terminal, 30...generation AI server, 31...generation AI model

Claims

1. An information processing device comprising: an acquisition unit that acquires a question from a first user to a second user, background information related to the background of the question, and role information related to the role of at least one of the first user and the second user; a determination unit that determines external information to refer to based on the background information and the role information; and an output unit that outputs input information to instruct a generative AI model to generate an answer sentence to the question based on the external information.

2. The information processing device described in claim 1, wherein the acquisition unit acquires the question including a demonstrative pronoun, and the output unit outputs the input information including content instructing that the estimated result of the information indicated by the demonstrative pronoun be included in the answer sentence.

3. The information processing device according to claim 1, wherein the background information is information relating to the content of the conversation that took place between the first user and the second user before the first user asked the question.

4. The information processing device according to claim 1, wherein the role information includes information relating to the roles of both the first user and the second user.

5. The information processing device according to claim 1, wherein the determination unit determines mode information indicating the source of external information to be referenced based on the background information and the role information, and determines the external information from among the information included in the source indicated by the mode information.

6. The information processing device described in claim 5, wherein the acquisition unit further acquires authority information regarding the access authority of at least one of the first user and the second user, and the determination unit determines the external information based on the mode information and the authority information.

7. The information processing device described in claim 5, wherein the determination unit inputs instruction information to a generative AI model, the instruction information including at least the background information, the role information, and content instructing to output the mode information corresponding to the background information and the role information, and determines the mode information based on the output result from the generative AI model.

8. The information processing device according to claim 5, wherein the output unit outputs the input information including content instructing that the answer sentence be generated based on predetermined answer sentence generation rules according to the mode information.

9. An information processing device as described in claim 8, wherein the external information includes a plurality of items each associated with information regarding importance, and the answer generation rule includes content that instructs taking into consideration the importance of each of the plurality of items.

10. An information processing method comprising the steps of: acquiring a question from a first user to a second user, background information relating to the background of the question, and role information relating to the role of at least one of the first user and the second user; determining external information to refer to based on the background information and the role information; and outputting input information to instruct a generative AI model to generate an answer to the question based on the external information.

Images