Information processing apparatus

Abstract

The objective of this invention is to automatically regenerate an answer when an inappropriate answer has been generated. The information processing apparatus includes: a first extraction unit that extracts a first summary feature quantity as a feature quantity of a summary of a query input by a user and a first keyword feature quantity as a feature quantity of keywords included in the query; a second extraction unit that extracts a second summary feature quantity as a feature quantity of a summary of an answer statement generated by a model for the query and a second keyword feature quantity as a feature quantity of keywords included in the answer statement; a determination unit that determines whether the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity exceeds a predetermined threshold; and an instruction unit that, if the distance exceeds the predetermined threshold, instructs the model to regenerate the answer statement.

Description

Technical Field

[0001] This invention relates to the technical field of an information processing device. Background Technology

[0002] As such a device, there is a known apparatus that outputs a response to a query from a user. For example, Patent Document 1 discloses a technique for automatically analyzing query emails from a mobile phone and automatically creating and sending response statements based on the analysis results.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2007-133463 Summary of the Invention

[0004] When using machine learning models to generate answers to user queries, inappropriate responses are sometimes produced. In such cases, multiple interactions occur between the user and the model (e.g., re-entering the query), creating technical issues that increase the time or workload required to obtain an appropriate answer.

[0005] The present invention was made in view of the above-mentioned problems, and its object is to provide an information processing device that can automatically regenerate a response when an inappropriate response is generated.

[0006] An information processing apparatus according to one aspect of the present invention comprises: a first extraction unit that extracts a first summary feature quantity as a feature quantity of a summary of a query statement input by a user and a first keyword feature quantity as a feature quantity of keywords contained in the query statement; a second extraction unit that extracts a second summary feature quantity as a feature quantity of a summary of an answer statement generated by a model for the query statement and a second keyword feature quantity as a feature quantity of keywords contained in the answer statement; a determination unit that determines whether the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity exceeds a predetermined threshold; and an indication unit that, if the distance exceeds the predetermined threshold, instructs the model to regenerate the answer statement. Attached Figure Description

[0007] Figure 1 This is a block diagram illustrating the hardware structure of the information processing device involved in the implementation method.

[0008] Figure 2 This is a block diagram illustrating the functional structure of the information processing apparatus involved in the implementation method.

[0009] Figure 3 It is a flowchart illustrating the operation of the information processing apparatus involved in the implementation method. Detailed Implementation

[0010] Hereinafter, embodiments of the information processing apparatus will be described with reference to the accompanying drawings.

[0011] (Hardware structure)

[0012] First, refer to Figure 1 The hardware structure of the information processing device involved in the implementation method will be described. Figure 1 This is a block diagram illustrating the hardware structure of the information processing device involved in the implementation method.

[0013] exist Figure 1 In this embodiment, the information processing apparatus 10 is configured to include an arithmetic unit 110, a storage unit 120, a communication unit 130, an input unit 140, and an output unit 150. The arithmetic unit 110, the storage unit 120, the communication unit 130, the input unit 140, and the output unit 150 are interconnected via a data bus.

[0014] The arithmetic processing unit 110 is configured to perform various arithmetic operations in the information processing unit 10. The arithmetic processing unit 110 may have a processor. The arithmetic processing unit 110 may have a single processor or multiple processors. That is, the arithmetic processing unit 110 may have more than one processor. Furthermore, the processor may be a multi-core processor. In the case where the arithmetic processing unit 110 has a single processor that functions as a multi-core processor, it can be said that the arithmetic processing unit 110 logically has multiple processors.

[0015] The processor included in the computing device 110 may be at least one of a central processing unit (CPU), a graphics processing unit (GPU), a field-programmable gate array (FPGA), and a tensor processing unit (TPU).

[0016] Storage device 120 may be at least one of random access memory (RAM), read-only memory (ROM), hard disk drive, magneto-optical disk drive, solid-state drive (SSD), and optical disk array. That is, storage device 120 may be implemented by a single device or by multiple devices.

[0017] Storage device 120 is capable of storing desired data. The computer program CP executed by the arithmetic unit 110 can be stored in storage device 120. When the arithmetic unit 110 executes the computer program CP, storage device 120 can temporarily store data temporarily used by the arithmetic unit 110.

[0018] Furthermore, the computer program CP can be recorded on a computer-readable and non-transitory recording medium. In this case, the computer program CP can be stored in the storage device 120 by reading the recording medium using a recording medium reading device (not shown) included in the information processing device 10. Additionally, at least one of optical discs, magnetic media, magneto-optical discs, semiconductor memory, and any medium capable of storing other programs can be used as the recording medium. The computer program CP can also be obtained from an external (not shown) device outside the information processing device 10 via the communication device 130. In other words, the computer program CP can be downloaded from an external device to the storage device 120 of the information processing device 10.

[0019] The arithmetic unit 110 (e.g., a processor) can perform the processing to be performed by the information processing unit 10 together with the storage device 120 storing the computer program CP (in other words, together with the storage device 120 and the computer program CP stored in the storage device 120). For example, the arithmetic unit 110 can implement logical function blocks for performing the processing to be performed by the information processing unit 10 within the arithmetic unit 110 (e.g., within the processor) by executing the computer program CP.

[0020] The communication device 130 is configured to communicate with external devices of the information processing device 10. Furthermore, the communication device 130 can perform both wired and wireless communication.

[0021] Input device 140 is a device capable of accepting information input from an external source to information processing device 10. Input device 140 may include user-operable devices (e.g., keyboard, mouse, touch panel, etc.) of information processing device 10. Input device 140 may include, for example, a recording medium reading device capable of reading information recorded on a recording medium removable from information processing device 10, such as a Universal Serial Bus (USB) memory. Furthermore, when information is input to information processing device 10 via communication device 130 (in other words, when information processing device 10 obtains information via communication device 130), communication device 130 can function as an input device.

[0022] Output device 150 is a device capable of outputting information to the outside of information processing device 10. Output device 150 may include a display device capable of outputting visual information such as characters and images. Output device 150 may also include a speaker capable of outputting auditory information such as voice. Output device 150 may be configured to output the aforementioned information to other devices (e.g., control information of other devices). Output device 150 may, for example, output information to a recording medium removable from information processing device 10, such as a USB memory. Furthermore, when information processing device 10 outputs information via communication device 130, communication device 130 may function as an output device.

[0023] <Functional Structure>

[0024] Next, refer to Figure 2 The functional structure of the information processing device 10 involved in the implementation method will be explained. Figure 2 This is a block diagram illustrating the functional structure of the information processing apparatus involved in the implementation method.

[0025] exist Figure 2 In this embodiment, the information processing apparatus 10 is configured to instruct the regeneration of an answer statement when the answer statement generated for a user-input query statement is inappropriate. The information processing apparatus 10 includes a query statement acquisition unit 210, a first extraction unit 220, a query statement output unit 230, an answer statement acquisition unit 240, a second extraction unit 250, a determination unit 260, a regeneration instruction unit 270, and an answer statement output unit 280 as components for implementing its function. Furthermore, the query statement acquisition unit 210, the first extraction unit 220, the query statement output unit 230, the answer statement acquisition unit 240, the second extraction unit 250, the determination unit 260, the regeneration instruction unit 270, and the answer statement output unit 280 can each be a processing block implemented by the aforementioned processing unit 110.

[0026] The query statement acquisition unit 210 is configured to acquire query statements input by the user. For example, the query statement acquisition unit 210 can acquire text input by the user using a keyboard or the like as a query statement. Alternatively, the query statement acquisition unit 210 can acquire speech input by the user using a microphone or the like as a query statement. In this case, the query statement acquisition unit 210 may have a speech recognition function that converts speech data into text data. The query statement acquired by the query statement acquisition unit 210 is configured to be output to the first extraction unit 220 and the query statement output unit 230.

[0027] The first extraction unit 220 is configured to extract a first summary feature and a first keyword feature from the query statement obtained by the query statement acquisition unit 210. The first summary feature is a feature of the summary of the query statement. The first keyword feature is a feature of the keywords contained in the query statement. The first summary feature and the first keyword feature can be extracted as feature vectors representing features. Furthermore, the method for extracting the first summary feature and the first keyword feature is not particularly limited. The first extraction unit 220 can, for example, use a machine learning model to extract the first summary feature and the first keyword feature. The first summary feature and the first keyword feature extracted by the first extraction unit 220 are each configured to be output to the determination unit 260.

[0028] The query statement output unit 230 is configured to output the query statement acquired by the query statement acquisition unit 210 to the answer generation model 300. The answer generation model 300 is a model for generating answer statements for the query statement. The answer generation model can be, for example, a Large Language Model (LLM). Furthermore, in Figure 2 The example given is that the answer generation model 300 is set outside the information processing device 10, but the information processing device 10 may also be configured to have the answer generation model 300.

[0029] The answer statement acquisition unit 240 is configured to acquire the answer statement generated by the answer generation model 300. The answer statement acquired by the answer statement acquisition unit 240 is configured to be output to the second extraction unit 250. Furthermore, the answer statement acquisition unit 240 is configured to output the answer statement to the answer statement output unit 280 based on the determination result in the determination unit 260.

[0030] The second extraction unit 250 is configured to extract a second summary feature and a second keyword feature from the answer statement acquired by the answer statement acquisition unit 240. The second summary feature is a feature of the summary of the answer statement. The second keyword feature is a feature of the keywords contained in the answer statement. The second summary feature and the second keyword feature can be extracted as feature vectors representing features. Furthermore, the method for extracting the second summary feature and the second keyword feature is not particularly limited. The second extraction unit 250 can, for example, use a machine learning model to extract the second summary feature and the second keyword feature. The second summary feature and the second keyword feature extracted by the second extraction unit 250 are each configured to be output to the determination unit 260.

[0031] The determination unit 260 is configured to calculate the distance between the first abstract feature and the first keyword feature and the second abstract feature and the second keyword feature. This distance can be a value representing the similarity or relevance between the first abstract feature and the first keyword feature and the second abstract feature and the second keyword feature. In this case, it can be said that the larger the calculated distance, the lower the similarity or relevance between the first abstract feature and the first keyword feature and the second abstract feature and the second keyword feature. The determination unit 260 can calculate the distance between the first abstract feature and the second abstract feature and the distance between the first keyword feature and the second keyword feature separately. Alternatively, the determination unit 260 can also calculate the distance between the first abstract feature and the first keyword feature and the second abstract feature and the second keyword feature together. For example, the determination unit 260 can calculate the distance between a feature formed by combining the first abstract feature and the first keyword feature and a feature formed by combining the second abstract feature and the second keyword feature.

[0032] The determination unit 260 is further configured to determine whether the calculated distance exceeds a predetermined threshold. This predetermined threshold is used to determine whether the answer generated by the answer generation model is appropriate for the query entered by the user. For example, if the distance between the first summary feature and the first keyword feature and the second summary feature and the second keyword feature exceeds the predetermined threshold, it can be determined that the similarity or relevance between the query and the answer is low (i.e., the answer is inappropriate). On the other hand, if the distance between the first summary feature and the first keyword feature and the second summary feature and the second keyword feature does not exceed the predetermined threshold, it can be determined that the similarity or relevance between the query and the answer is high (i.e., the answer is appropriate). The determination result performed by the determination unit 260 is configured to be output to the answer acquisition unit 240 and the regeneration instruction unit 270.

[0033] The regeneration instruction unit 270 is configured to output an instruction to the answer generation model 300 to regenerate the answer statement. Specifically, if the distance calculated by the determination unit 260 exceeds a predetermined threshold, the regeneration instruction unit 270 outputs an instruction to regenerate the answer statement. For example, the regeneration instruction unit 270 can instruct the regeneration of the answer statement by modifying a portion of the query statement entered by the user and outputting it to the answer generation model 300.

[0034] The response statement output unit 280 is configured to output a response statement to the user who has entered a query statement. Specifically, the response statement output unit 280 outputs a response statement to the user if the distance calculated by the determination unit 260 does not exceed a predetermined threshold.

[0035] (The flow of the action)

[0036] Next, refer to Figure 3 The operation flow of the information processing device 10 according to the embodiment will be explained. Figure 3 It is a flowchart illustrating the operation of the information processing apparatus involved in the implementation method.

[0037] like Figure 3 As shown, when the information processing apparatus 10 according to the embodiment begins operation, the query statement acquisition unit 210 first acquires the query statement input by the user (step S101). Then, the first extraction unit 220 extracts the first summary feature and the first keyword feature from the query statement acquired by the query statement acquisition unit 210 (step S102). Furthermore, the query statement output unit 230 outputs the query statement acquired by the query statement acquisition unit 210 to the answer generation model 300 (step S103). As a result, an answer statement for the query statement is generated in the answer generation model 300.

[0038] Next, the answer statement acquisition unit 240 acquires the answer statement generated in the answer generation model 300 (step S104). Then, the second extraction unit 250 extracts the second summary feature and the second keyword feature from the answer statement acquired by the answer statement acquisition unit 240 (step S105).

[0039] Next, the determination unit 260 calculates the distance between the first abstract feature and the first keyword feature and the second abstract feature and the second keyword feature, and determines whether the calculated distance exceeds a predetermined threshold (step S106). Alternatively, if the distance calculated by the determination unit 260 does not exceed the predetermined threshold (step S106: No), the response statement output unit 280 outputs the response statement acquired by the response statement acquisition unit 240 to the user (step S107). In this case, the regeneration of the response statement may not be performed.

[0040] On the other hand, if the distance calculated by the determination unit 260 exceeds a predetermined threshold (step S106: Yes), the regeneration instruction unit 270 outputs an instruction to regenerate the answer to the answer generation model 300 (step S108). As a result, a new answer statement for the query statement is generated in the answer generation model 300.

[0041] Furthermore, if the answer statement is regenerated by the answer generation model 300, the processing from step S104 is repeated. This process is repeated until the distance calculated by the determination unit 260 does not exceed a predetermined threshold (step S106: No), at which point the answer statement output unit 280 outputs the answer statement obtained by the answer statement acquisition unit 240 (i.e., the regenerated answer statement) to the user (step S107).

[0042] (Technical effect)

[0043] Next, the technical effects obtained by the information processing apparatus 10 according to the implementation method will be explained.

[0044] like Figures 1 to 3 As explained, in the information processing apparatus 10 according to the embodiment, it is determined whether the distance between the first summary feature and the first keyword feature extracted from the query statement and the second summary feature and the second keyword feature extracted from the answer statement exceeds a predetermined threshold. Therefore, it is possible to automatically determine whether the answer statement is appropriate for the query statement. Furthermore, in the information processing apparatus 10 according to the embodiment, if the distance calculated by the determination unit 260 exceeds the predetermined threshold, an instruction to regenerate the answer is output to the answer generation model 300. Therefore, even if the answer statement generated by the answer generation model 300 is inappropriate, the regeneration of the answer statement is repeatedly performed until it becomes appropriate. As a result, an appropriate answer statement can be output to the user.

[0045] Furthermore, if the appropriateness of the response cannot be automatically determined, the user may need to revise and re-enter the query. In this case, the user's workload or the time required to obtain an appropriate response increases, potentially leading to distrust of the response generation model 300. However, in this embodiment, as described above, the appropriateness of the response can be automatically determined. Therefore, the aforementioned inconveniences can be avoided, and an appropriate response can be output to the user.

[0046] The invention described below is an embodiment derived from the above-described implementation.

[0047] An information processing apparatus according to one aspect of the present invention includes: a first extraction unit that extracts a first summary feature quantity as a feature quantity of a query statement input by a user and a first keyword feature quantity as a feature quantity of keywords contained in the query statement; a second extraction unit that extracts a second summary feature quantity as a feature quantity of a response statement generated by a model for the query statement and a second keyword feature quantity as a feature quantity of keywords contained in the response statement; a determination unit that determines whether the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity exceeds a predetermined threshold; and an indication unit that, if the distance exceeds the predetermined threshold, instructs the model to regenerate the response statement. In the above embodiment, "first extraction unit 220" is equivalent to "first extraction unit" in one example, "second extraction unit 250" is equivalent to "second extraction unit" in another example, "determination unit 260" is equivalent to "determination unit" in another example, and "regeneration indication unit 270" is equivalent to "indication unit" in another example.

[0048] In the information processing apparatus described above, an output unit may also be included, which outputs the regenerated answer statement to the user when the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity extracted from the regenerated answer statement is below the predetermined threshold. In this way, the answer statement can be output to the user if an appropriate answer statement has been generated (for example, if the answer statement is made appropriate through regeneration). In the above embodiment, "answer statement output unit 280" is an example of "output unit".

[0049] The present invention is not limited to the embodiments described above, and appropriate modifications can be made without departing from the spirit or idea of ​​the invention as read in its entirety from the claims and description. Information processing apparatuses that accompany such modifications are also included within the technical scope of the present invention.

[0050] Symbol Explanation

[0051] 10 - Information processing device; 110 - Computational device; 120 - Storage device; 130 - Communication device; 140 - Input device; 150 - Output device; 210 - Query statement acquisition unit; 220 - First extraction unit; 230 - Query statement output unit; 240 - Response statement acquisition unit; 250 - Second extraction unit; 260 - Judgment unit; 270 - Regeneration instruction unit; 280 - Response statement output unit; 300 - Response generation model.

Claims

1. An information processing device, characterized in that, have: The first extraction unit extracts a first summary feature quantity, which is a feature quantity of a summary of a query statement input by the user, and a first keyword feature quantity, which is a feature quantity of keywords contained in the query statement. The second extraction unit extracts a second summary feature quantity as a feature quantity of the answer statement generated by the model for the query statement and a second keyword feature quantity as a feature quantity of the keywords contained in the answer statement. The determination unit determines whether the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity exceeds a predetermined threshold. and An instruction unit that, if the distance exceeds a predetermined threshold, instructs the model to regenerate the response statement.

2. The information processing device according to claim 1, characterized in that, It also includes an output unit that outputs the regenerated answer statement to the user when the distance between the first summary feature quantity and the first keyword feature quantity and the second summary feature quantity and the second keyword feature quantity extracted from the regenerated answer statement is below the predetermined threshold.

Images