Information processing device, information processing method, and information processing program

Abstract

To improve the accuracy of response information to questions containing unknown words.SOLUTION: An information processing device comprises a storage unit which stores a first learning model for outputting the meanings of words and a second learning model for outputting response information to questions, and a processor. The processor performs first identification processing for using the first learning model to identify the meanings of words in a question sentence, analysis processing for analyzing the meaning of the question sentence on the basis of the identified meanings of the words, and acquisition processing for inputting the analysis results of the question sentence into the second learning model to acquire the first response information. The analysis processing further includes outputting inquiry information asking for the meanings of words which could not be identified, identifying the meanings of words which could not be identified on the basis of the second response information to the inquiry information, and analyzing the meaning of the question sentence on the basis of the meanings of the words identified on the basis of the first identification processing and the second response information.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

Background Art

[0002] In recent years, artificial intelligence (also referred to as "AI") that generates content such as sentences and images in response to a prompt has been used. For example, in a technique for generating a sentence from keywords included in an input sentence, when the input sentence includes an unknown word that is not in the learned dictionary, a technique has been proposed in which the unknown word in the input sentence is replaced with a predicted word to generate a sentence, and the predicted word in the generated sentence is replaced with the unknown word (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] When generating response information for a question sentence from a user, if the question sentence includes an unknown word that is not in the learned dictionary, the accuracy of the response information for the question sentence may decrease. For example, even when the unknown word included in the input sentence is replaced with a predicted word, it is conceivable that the predicted word is different from the meaning intended by the user. Even in such a case, the accuracy of the generated response information decreases.

[0005] One aspect of the disclosed technology aims to provide an information processing apparatus, an information processing method, and an information processing program capable of improving the accuracy of response information for input information such as a question sentence. In particular, an object is to provide an information processing apparatus, an information processing method, and an information processing program that can output relevant response information as an output for the input information based on the meaning and content of the input information.

Means for Solving the Problem

[0006] One aspect of the disclosed technology is exemplified by the following information processing apparatus. The present information processing apparatus includes a storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs response information for the question when the question is input, and a processor connected to the storage unit. The processor performs a first specifying process of specifying the meaning of each word included in the question sentence input by the user using the first learning model, an analyzing process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, an inputting process of inputting the analysis result of the question sentence by the analyzing process into the second learning model, an obtaining process of obtaining first response information for the question sentence from the second learning model, and an outputting process of outputting the first response information. The analyzing process outputs inquiry information for inquiring about the meaning of the word when the meaning cannot be specified in the first specifying process, or when the meaning of the word can be specified but the first response information cannot be output by the outputting process based on the meaning of the word, receives second response information for the inquiry information, and when the first specifying process specifies the meaning of the word that could not be specified based on the second response information, analyzes the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second response information.

[0007] When the above-described inquiry text contains words whose meanings cannot be specified, this information processing apparatus outputs inquiry information. Then, this information processing apparatus specifies the meanings of the words whose meanings could not be specified based on the above-described second response information for the inquiry information, and analyzes the meaning of the inquiry text. Through such an exchange of inquiry information and second response information, this information processing apparatus can analyze the meaning of the inquiry text from the user with higher accuracy. As a result, this information processing apparatus can improve the accuracy of the response information for an inquiry text containing unknown words.

[0008] This information processing apparatus may further have the following features. The above-described processor further executes a learning process of training the first learning model based on the word specified based on the above-described second response information and the meaning of the word specified based on the above-described second response information. By having such a feature, this information processing apparatus can expand the words that can be specified by the first learning model. As a result, this information processing apparatus can reduce the workload of causing the user to provide a response to the above-described inquiry information.

[0009] This information processing apparatus may further have the following features. The above-described processor further executes a storage process of storing the above-described second response information in the above-described storage unit in association with the above-described inquiry text and the word whose meaning could not be specified. By having such a feature, this information processing apparatus can specify the meaning of the word that could not be specified by the first learning model by referring to the above-described storage unit. As a result, this information processing apparatus can reduce the workload of causing the user to provide a response to the above-described inquiry information.

[0010] This information processing apparatus may further have the following features. Specifying based on the above-described second response information further includes notifying that a response is not possible when the meaning of the word whose meaning could not be specified cannot be specified based on the above-described second response information. By having such a feature, this information processing apparatus can suppress the output of a response that does not conform to the intention of the inquiry text.

[0011] The information processing apparatus may further have the following features. The output processing further includes a process of including location information indicating the location of materials related to the first response information in the first response information. By having such a feature, the information processing apparatus can present materials that are useful as response information to a question to the user.

[0012] The information processing apparatus may further have the following features. The processor executes the input processing of inputting the first analysis result for the first question sentence into the second learning model, and the acquisition processing of acquiring the first response information for the second question sentence whose analysis result has already been input into the second learning model from the second learning model. By having such a feature, the information processing apparatus can shorten the time until the generation of the final response information for the question sentence from the user.

[0013] The disclosed technology can also be understood from the aspects of an information processing method and an information processing program.

Advantages of the Invention

[0014] According to the disclosed technology, the accuracy of response information for a question sentence containing unknown words can be improved.

Brief Description of the Drawings

[0015]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

DETAILED DESCRIPTION OF THE INVENTION

[0016] <Embodiment> Hereinafter, embodiments will be described with reference to the drawings. FIG. 1 is a diagram showing an example of an answering system 1 according to an embodiment. The answering system 1 includes an answering device 2, a user terminal 3, and a network N1. The answering device 2 and the network N1 are communicably connected to each other by the network N1.

[0017] When the answering device 2 receives a question sentence input to the user terminal 3 by the user U1 from the user terminal 3, it is an information processing device that transmits an answer sentence for the question sentence to the user terminal 3. The question sentence is an example of input information, and is, for example, a sentence indicating the content for which the user U1 requests an answer from the answering device 2, includes one or more words, and is transmitted from the user terminal 3 to the answering device 2. Question information including data other than the question sentence and the sentence may be transmitted from the user terminal 3 to the answering device 2. Examples of the data other than the sentence included in the answer information include audio data, video data, image data, and the like. The user terminal 3 is an information processing device that transmits a question sentence input by the user U1 to the answering device 2 and outputs an answer sentence received from the answering device 2 to a display device such as a display.

[0018] The network N1 communicably connects information processing devices to each other. The network N1 is, for example, a Local Area Network (LAN), a mobile communication system, or the Internet. The network N1 may be wired or wireless.

[0019] FIG. 2 is a diagram showing an example of the hardware configuration of the answering device 2 according to the embodiment. The answering device 2 includes a CPU 21, a main memory unit 22, an auxiliary storage unit 23, a communication unit 24, and a connection bus B2. The CPU 21, the main memory unit 22, the auxiliary storage unit 23, and the communication unit 24 are connected to each other by the connection bus B2.

[0020] The CPU 21 is also called a microprocessor unit (MPU) or a processor. The CPU 21 is not limited to a single processor and may have a multi-processor configuration. Also, a single CPU 21 connected by a single socket may have a multi-core configuration. At least a part of the processing executed by the CPU 21 may be performed by a processor other than the CPU 21, for example, a dedicated processor such as a Digital Signal Processor (DSP), a Graphics Processing Unit (GPU), a numerical calculation processor, a vector processor, an image processing processor, a Neural Processing Unit (NPU), etc. Also, at least a part of the processing executed by the CPU 21 may be executed by an integrated circuit (IC) or other digital circuits. Also, an analog circuit may be included in at least a part of the CPU 21. The integrated circuit includes a Large Scale Integrated circuit (LSI), an Application Specific Integrated Circuit (ASIC), and a programmable logic device (PLD). The PLD includes, for example, a Field-Programmable Gate Array (FPGA). The CPU 21 may be a combination of a processor and an integrated circuit. The combination is called, for example, a microcontroller unit (MCU), a System-on-a-chip (SoC), a system LSI, a chipset, etc. In the answering device 2, the CPU 21 expands the program stored in the auxiliary storage unit 23 into the working area of the main storage unit 22 and controls the peripheral devices through the execution of the program. Thereby, the answering device 2 can execute processing that meets a predetermined purpose. The main storage unit 22 and the auxiliary storage unit 23 are recording media readable by the CPU 21.

[0021] The main storage unit 22 is exemplified as a storage unit directly accessible from the CPU 21. The main storage unit 22 includes a Random Access Memory (RAM) and a Read Only Memory (ROM).

[0022] The auxiliary storage unit 23 reads and writes various programs and various data to and from a recording medium. The auxiliary storage unit 23 is also called an external storage device. The auxiliary storage unit 23 stores an operating system (OS), various programs, various tables, etc. The OS includes a communication interface program for transferring data with external devices connected via the communication unit 24. The external devices include, for example, other information processing devices and external storage devices connected by a computer network or the like. Note that the auxiliary storage unit 23 may be a part of a cloud system which is a group of computers on a network, for example.

[0023] The auxiliary storage unit 23 is, for example, an Erasable Programmable ROM (EPROM), a Solid State Drive (SSD), a Hard Disk Drive (HDD), etc. Also, the auxiliary storage unit 23 is, for example, a Compact Disc (CD) drive device, a Digital Versatile Disc (DVD) drive device, a Blu-ray (registered trademark) Disc (BD) drive device, etc. Also, the auxiliary storage unit 23 may be provided by a Network Attached Storage (NAS) or a Storage Area Network (SAN).

[0024] The communication unit 24 is, for example, an interface with the network N1. The communication unit 24 communicates with external devices via the network N1.

[0025] FIG. 3 is a diagram showing an example of the hardware configuration of the user terminal 3 according to the embodiment. The user terminal 3 includes a CPU 31, a main memory unit 32, an auxiliary storage unit 33, a communication unit 34, an input unit 35, an output unit 36, and a connection bus B3. Since the CPU 31, the main memory unit 32, the auxiliary storage unit 33, the communication unit 34, and the connection bus B3 have the same configuration as the CPU 21, the main memory unit 22, the auxiliary storage unit 23, the communication unit 24, and the connection bus B2 of the answering device 2, their description will be omitted.

[0026] The input unit 35 receives input from the user U1. The input unit 35 is, for example, a keyboard, a mouse, a trackball, a touch panel, a voice input device, or the like. The output unit 36 outputs data processed by the CPU 31 and data stored in the main memory unit 32. The output unit 36 can include, for example, a display, a printer, and a speaker.

[0027] <Processing blocks of the answering device 2> FIG. 4 is a diagram showing an example of the processing blocks of the answering device 2 according to the embodiment. The answering device 2 includes a login processing unit 200, a first analysis unit 201, a confirmation unit 202, an unknown word identification unit 203, a second analysis unit 204, a selection unit 205, an analysis result input unit 206, an answer acquisition unit 207, an output unit 208, a charging unit 209, a management database 210 (described as "management DB 210" in the figure), a word model 220, and an answer model 230. The answering device 2 executes, by the CPU 21 executing a computer program expandably deployed in the main memory unit 22, the processing of each part of the answering device 2 such as the login processing unit 200, the first analysis unit 201, the confirmation unit 202, the unknown word identification unit 203, the second analysis unit 204, the selection unit 205, the analysis result input unit 206, the answer acquisition unit 207, the output unit 208, the charging unit 209, the management database 210, the word model 220, and the answer model 230.

[0028] The word model 220 outputs text data indicating the meaning of the input word. The word model 220 is a learning model constructed, for example, by machine learning using words and text data indicating the meaning of the words as teacher data. The word model 220 may be, for example, a table associating words with text data indicating the meaning of the words. The word model 220 is prepared for each language such as Japanese, English, German, etc. The word model 220 is an example of the "first learning model".

[0029] The answer model 230 generates an answer sentence for the input question sentence. The answer sentence is a sentence indicating an answer to the question sentence. The answer model 230 may generate answer information including an answer sentence and data other than text for the input question sentence. Examples of data other than text included in the answer information include audio data, video data, image data, etc. The answer model 230 generates an answer sentence using, for example, a large language model (LLM). The answer sentence generated by the answer model 230 may include information indicating the location of materials related to the answer sentence for the question sentence. The information indicating the location of the materials is, for example, a Uniform Resource Identifier (URI). Also, when the conditions included in the input question sentence are insufficient for outputting an answer sentence, the answer model 230 may output to that effect. The answer model 230 includes, for example, a plurality of learning models according to the information field of the question sentence. The answer model 230 is a learning model constructed based on teacher data, for example. The answer model 230 is an example of the "second learning model". The answer sentence generated by the answer model 230 is an example of the "answer information". The information indicating the location of the materials is an example of the "location information".

[0030] FIG. 5 is a diagram showing an example of the answer model 230. The answer model 230 has, for example, a first maintenance model 231, a second maintenance model 232, a third maintenance model 233, a first travel model 234, and a second travel model 235. The first maintenance model 231, the second maintenance model 232, and the third maintenance model 233 are a type of learning model constructed using information related to vehicle maintenance as teacher data. The first travel model 234 and the second travel model 235 are a type of learning model constructed using information related to travel as teacher data. That is, the answer model 230 is composed of multiple types of learning models, and the types include multiple learning models that differ according to the information field of the input information (input sentence, question sentence). The learning models at this time may also be referred to as the first type of learning model and the second type of learning model according to their types. Hereinafter, as an example of the first type of learning model, the first maintenance model 231, the second maintenance model 232, and the third maintenance model 233 are shown, and as an example of the second type of learning model, the first travel model 234 and the second travel model 235 are shown. Examples of other learning models will be described later.

[0031] The first maintenance model 231 is constructed using information related to automobile maintenance as teacher data. Examples of information related to automobile maintenance include, for example, maintenance manuals and official websites published by automobile manufacturers, information related to maintenance work published by automobile maintenance factories, information related to maintenance published by automobile mechanics, blogs and social networking services (SNS) related to maintenance cases published by automobile mechanics, information published in automobile information magazines, and custom cases of automobiles posted on Internet bulletin boards, etc.

[0032] The second maintenance model 232 is constructed using highly reliable information among information related to automobile maintenance as teacher data. Examples of highly reliable information include, for example, maintenance manuals and official websites published by automobile manufacturers, information related to maintenance work published by automobile maintenance factories, information related to maintenance published by automobile mechanics, etc.

[0033] The third maintenance model 233 is constructed using information related to vehicle maintenance as training data. The third maintenance model 233 aims to build a lightweight model and is constructed, for example, with less training data than the first maintenance model 231.

[0034] Since the training data used for constructing the second maintenance model 232 is limited to highly reliable information among information related to vehicle maintenance, there is a tendency for the fields in which answers can be generated to be narrower than those of the first maintenance model 231. On the other hand, the second maintenance model 232 constructed based on such highly reliable information tends to generate more reliable answers to questions related to the maintenance of the input vehicle than the first maintenance model 231. In other words, although the reliability of the answers generated by the first maintenance model 231 is inferior to that of the second maintenance model 232, it tends to be able to generate answers to a wide range of questions related to maintenance. Also, since the third maintenance model 233 is a lightweight model, it can generate answers faster than the first maintenance model 231.

[0035] The first travel model 234 is constructed using information related to travel as training data. Examples of information related to travel include tourism information published by prefectural governments, overseas safety information published by the Ministry of Foreign Affairs, tourism information published by government agencies responsible for tourism promotion in each country, websites where travel agencies publish information related to tourist destinations, blogs and SNS where individuals publish records of their travels, etc.

[0036] The second travel model 235 is constructed using highly reliable information among information related to travel as training data. Examples of highly reliable information include, for example, information related to travel.

[0037] The training data used for constructing the second travel model 235 is highly reliable among information related to travel Since it is limited to highly reliable information, there is a tendency for the field in which answers can be generated to be narrower than that of the first travel model 234. On the other hand, the second travel model 235 constructed based on such highly reliable information tends to generate more reliable answers to questions regarding the input travel than the first travel model 234. In other words, although the reliability of the answers generated by the first travel model 234 is inferior to that of the second travel model 235, it tends to be able to generate answers to a wide range of questions regarding travel.

[0038] In addition, in FIG. 5, as learning models of the answer model 230, the first maintenance model 231 and the second maintenance model 232 related to automobile maintenance, the first travel model 234 and the second travel model 235 related to travel are illustrated. However, the answer model 230 may further have learning models other than these. Further, the answer model 230 may have other learning models instead of the first maintenance model 231, the second maintenance model 232, the first travel model 234, and the second travel model 235. The answer model 230 may have, for example, a learning model that outputs text data indicating information related to health maintenance, a learning model that outputs text data indicating information related to movies, a learning model that outputs text data indicating information related to music, and the like.

[0039] Note that, although details will be described later, for a single input information such as an input sentence or a question sentence, a single answer model is not used. Instead, it may be divided into a plurality of input contents (input information in a broad sense) according to the delimiters of the input information of the input sentence or the question sentence (for example, punctuation marks, line breaks, etc.), and the answer model 230 may be used for each separated input content.

[0040] Returning to FIG. 4, the management database 210 is a database that manages the features of each model that the answer model 230 has. FIG. 6 is a diagram showing an example of the model feature table 211 stored in the management database 210 in the embodiment. The model feature table 211 includes columns of "model", "label", "reliability", and "processing speed". In "model", information indicating each model that the answer model 230 has is stored. In "label", information indicating the field of the teacher data used for building the model is stored. In other words, it can be said that the "field suitable for the answer by the model" is stored in "label". Note that a label indicating a plurality of fields may be stored in "label". In "reliability", information indicating the reliability of the answer generated by the model is stored. In "processing speed", information indicating the speed from when a question is input until an answer is generated is stored. Note that the information stored in the model feature table 211 is not limited to the information illustrated in FIG. 6. The model feature table 211 may include, for example, the amount of teacher data used for model construction, information indicating the usage fee of the model, and the like. The information stored in "label", "reliability", and "processing speed" is an example of "feature information indicating features related to answer accuracy". The "label" of the model feature table 211 is an example of "first field information".

[0041] Returning to FIG. 4, the login processing unit 200 performs the login processing of the user U1. The login processing unit 200 performs, for example, login processing using a user name and a password.

[0042] When the first analysis unit 201 receives an input sentence exemplified by a question sentence from the user U1 and an answer sentence from the answer model 230, it generates a thread of the input sentence. The input sentence is an example of input information, includes one or more words, and is, for example, a sentence input to the first analysis unit 201 and targeted for analysis by the first analysis unit 201. Examples of the input sentence include a question sentence input by the user U1, an inquiry sentence output by the confirmation unit 202 described later, an answer sentence from the user U1 to the inquiry sentence, and the like. Note that input information including data other than the input sentence and the sentence may be input to the first analysis unit 201. Examples of data other than the sentence included in the input information include audio data, video data, image data, and the like. The thread includes a series of question sentences and answer sentences. The series of question sentences and answer sentences are, for example, a question sentence input by the user U1, an inquiry sentence generated by the confirmation unit 202 described later for the question sentence, an answer sentence input by the user U1 to the inquiry sentence, and an answer sentence acquired by the answer acquisition unit 207 described later. The thread may include an additional question sentence from the user U1 made for the answer sentence acquired by the answer acquisition unit 207. The first analysis unit 201 stores, for example, the generated thread in the auxiliary storage unit 23 in association with the user name of the user U1 for whom login processing has been performed by the login processing unit 200. The thread includes an inquiry sentence generated by the confirmation unit 202 for the question sentence, an answer sentence input by the user U1 to the inquiry sentence, and an answer sentence acquired by the answer acquisition unit 207 described later. The thread may include an additional question sentence from the user U1 made for the answer sentence acquired by the answer acquisition unit 207. The first analysis unit 201 stores, for example, the generated thread in the auxiliary storage unit 23 in association with the user name of the user U1 for whom login processing has been performed by the login processing unit 200.

[0043] In addition, the first analysis unit 201 identifies the language of the input sentence. The first analysis unit 201 identifies, for example, whether the input sentence is in Japanese, English, German, or the like. The first analysis unit 201 may include, for example, a learning model that outputs text data indicating the language used in the sentence when the sentence is input, and identify the language using the learning model.

[0044] Then, the first analysis unit 201 identifies the meaning of each word included in the input sentence. The first analysis unit 201 decomposes the input sentence into morphemes (e.g., words) by performing morphological analysis on the question sentence. The first analysis unit 201 may further identify the part of speech of each morpheme by morphological analysis. The first analysis unit 201 inputs each word included in the input sentence into the word model 220 to identify its meaning. Note that the first analysis unit 201 may select the word model 220 used for identifying the meaning of words according to the identified language of the input sentence. For example, if the identified language of the input sentence is Japanese, the first analysis unit 201 uses the Japanese word model 220 to identify the meaning of words. The first analysis unit 201 identifies a word whose meaning cannot be identified even when input into the word model 220 as an unknown word. In addition, the first analysis unit 201 may identify a word that cannot be normally clustered by the second analysis unit 204 described later as an unknown word. The processing by the first analysis unit 201 is an example of the "first identification processing".

[0045] The confirmation unit 202 generates an inquiry sentence for inquiring the meaning of the unknown word identified by the first analysis unit 201 to the user U1. That is, the confirmation unit 202 generates an inquiry sentence for inquiring the meaning of the word that could not be identified by the first analysis unit 201. For example, the confirmation unit 202 may store a stereotyped sentence such as "What does this word '●●' mean?" in the auxiliary storage unit 23 in advance, and generate an inquiry sentence by replacing "●●" with the unknown word. In addition, when an answer sentence indicating insufficient information is output from the answer model 230, the confirmation unit 202 may use the answer sentence as the inquiry sentence. The confirmation unit 202 outputs the inquiry sentence to the user terminal 3. The processing by the confirmation unit 202 is an example of the processing of "outputting inquiry information". The inquiry sentence is an example of the "inquiry information".

[0046] In addition, in the confirmation unit 202, in addition to generating the above inquiry sentence, while the meaning of a word can be specified by the first specific processing, even when it is not possible or difficult to output a response sentence (first response information) by the output processing described later based on the meaning of the word, it is also possible to generate an inquiry sentence for resolving elements (words, context, etc.) that are unclear in generating the response sentence. It is sufficient that at least one of these generated inquiry sentences is used.

[0047] When the unknown word specifying unit 203 receives a response sentence input by the user U1 for the inquiry sentence output by the confirmation unit 202, the received response sentence is analyzed. The unknown word specifying unit 203 causes, for example, the first analysis unit 201 to perform morphological analysis on the received response sentence and specify the meaning of the words included in the response sentence. The unknown word specifying unit 203 acquires the meaning of the response sentence using the result of specifying the meaning of the word by the first analysis unit 201. The unknown word specifying unit 203 specifies the meaning of the unknown word based on the meaning of the acquired response sentence. The response sentence input by the user U1 for the inquiry sentence is an example of "second response information".

[0048] The unknown word specifying unit 203 may perform additional learning processing on the word model 220 using the specified meaning of the unknown word and the correspondence relationship of the unknown word as teacher data. Further, the unknown word specifying unit 203 may associate the unknown word, the specified meaning of the unknown word, and the input sentence including the unknown word and store them in the auxiliary storage unit 23. Here, when the meaning of the unknown word cannot be specified even by the meaning of the response sentence acquired by the unknown word specifying unit 203, the confirmation unit 202 may be caused to output a further inquiry sentence. Further, when the meaning of the unknown word cannot be specified even by the meaning of the response sentence for the inquiry sentence, the unknown word specifying unit 203 may instruct the output unit 208 to output a message indicating that a response cannot be provided to the user terminal 3.

[0049] The second analysis unit 204 performs summary grasping of the input sentence based on morphological analysis, assigns (also referred to as setting) labels to the input sentence, and clustering. The second analysis unit 204 grasps, for example, an outline of what the input sentence is about based on the result of the morphological analysis by the first analysis unit 201. The second analysis unit 204 can grasp, for example, whether the input sentence is a question regarding travel or a question regarding automobile maintenance. Also, by performing clustering, an overall image of the content of the input sentence can be grasped, for example, according to "Grice's maxims".

[0050] The second analysis unit 204 analyzes the meaning of the input sentence, for example, using the meaning of the words specified by the first analysis unit 201 and the meaning of the unknown words specified by the unknown word specifying unit 203. In the analysis by the second analysis unit 204, for example, labels representing the meanings of these morphemes (words and sentences) decomposed by the morphological analysis by the first analysis unit 201 are assigned to each of the morphemes. For example, the meaning of the word specified by the word model 220 may be adopted as a label.

[0051] The second analysis unit 204 performs clustering of each morpheme included in the input sentence so that morphemes with the same label belong to the same cluster. Based on the result of the clustering, the second analysis unit 204 obtains the meaning of the entire question sentence. Examples of the meaning of the entire input sentence analyzed by the second analysis unit 204 include the field targeted by the input sentence, whether the input sentence has positive or negative content, whether it is a question sentence seeking highly reliable information, or whether it is a question sentence seeking information that can be used as a reference even with low reliability. The second analysis unit 204 assigns a label to the question sentence based on the obtained meaning of the question sentence. Further, the second analysis unit 204 may obtain the amount of text of the input sentence. The amount of text may be, for example, the number of characters included in the input sentence. Also, the amount of text may be the number of words included in the input sentence. Further, a case may occur where clustering cannot be performed normally as a result of the clustering by the second analysis unit 204. Examples of the case where clustering cannot be performed normally include a case where there is a cluster in which a small number (for example, one) of morphemes are classified. The label assigned to the question sentence is an example of "second field information".

[0052] Based on the analysis result by the second analysis unit 204, the selection unit 205 refers to the model feature table 211 and selects one or more answer models 230 to be used for answering the question sentence. For example, when it is analyzed by the second analysis unit 204 that the label assigned to the question sentence is "automobile maintenance", the selection unit 205 may select, as candidates for the learning models using the first maintenance model 231, the second maintenance model 232, and the third maintenance model 233 that include "automobile maintenance" in the "label" in the model feature table 211. The answer model 230 selected by the selection unit 205 is an example of the "selected learning model".

[0053] If it is analyzed by the second analysis unit 204 that the reliability of the answer sought by the question sentence is high, the selection unit 205 may select the highly reliable second maintenance model 232 among the first maintenance model 231, the second maintenance model 232, and the third maintenance model 233. Also, if it is analyzed by the second analysis unit 204 that a large amount of information that can be used as a reference is sought even with low reliability, the selection unit 205 When analyzed, all of the first maintenance model 231, the second maintenance model 232, and the third maintenance model 233 whose labels include "automobile maintenance" may be selected as the learning model used in the answer to the question sentence. The process of selecting one or more answer models 230 by the selection unit 205 is an example of the "selection process".

[0054] Here, the selection unit 205 calculates the degree of relevance between the label assigned to the question sentence by the second analysis unit 204 and the "label" in the model feature table 211, and preferentially selects the answer model 230 with a higher calculated degree of relevance, and may select the answer model 230 used in the answer to the question sentence. For example, the selection unit 205 may select the answer model 230 with the highest calculated degree of relevance as the answer model 230 used in the answer to the question sentence. Further, for example, the selection unit 205 may select one or more answer models 230 used in the answer to the question sentence in descending order of the calculated degree of relevance. There is no limitation on the method of calculating the degree of relevance. For example, the selection unit 205 may calculate the similarity between the label of the question sentence assigned by the second analysis unit 204 and the word (or sentence) indicating the label stored in the "label" of the model feature table 211, and use the calculated similarity as the degree of relevance. Various known techniques can be applied to the method of calculating the similarity of words (or sentences). The degree of relevance is an example of the "field relevance".

[0055] The analysis result input unit 206 inputs the analysis result by the second analysis unit 204 into the answer model 230 selected by the selection unit 205. When the analysis result input unit 206 receives question information from the user terminal 3, it may also input data other than the text included in the question information into the answer model 230. The analysis result input unit 206 may, for example, process the analysis result by the second analysis unit 204 so as to easily obtain an answer by the answer model 230, and input the processed analysis result into the answer model 230. In the processing by the analysis result input unit 206, for example, the analysis result is processed such that an instruction to the answer model 230 is placed at the head and supplementary information is placed thereafter. When a plurality of answer models 230 are selected by the selection unit 205, the analysis result input unit 206 inputs the analysis result by the second analysis unit 204 into each of the selected plurality of answer models 230. Further, when inputting the analysis result by the second analysis unit 204 into a plurality of answer models 230, the analysis result input unit 206 may, for example, input the analysis result by the second analysis unit 204 in order from the answer model 230 with a higher relevance calculated by the selection unit 205 to the answer model 230 with a lower relevance.

[0056] The answer acquisition unit 207 acquires the answer sentence generated by the answer model 230. When a plurality of answer models 230 are selected by the selection unit 205, the answer acquisition unit 207 acquires the answer sentences from the selected plurality of answer models 230. The answer acquisition unit 207 may process the answer sentences from the acquired answer models 230. Examples of the processing by the answer acquisition unit 207 include synthesis and summary of these plurality of answer sentences when answer sentences are acquired from a plurality of answer models 230.

[0057] As for the synthesis of the response sentences by the response acquisition unit 207, for example, methods such as directly concatenating the response sentences from a plurality of response models 230 and arranging and synthesizing the response sentences of the response models 230 with high relevance so as to give priority can be cited. As a method of giving priority to the response sentences of the response models 230 with high relevance, for example, a method of arranging the response sentences from the response models 230 with high relevance at the beginning of the response sentences to be output to the output unit 208, a method of setting the response sentences from the response models 230 with high relevance to a larger size than the response sentences from the response models 230 with low relevance, etc. can be cited. Further, when the response acquisition unit 207 acquires response sentences from a plurality of response models 230, these plurality of response sentences may be summarized. When the response acquisition unit 207 completes the processing of the response sentences, it instructs the output unit 208 to output the processed response sentences. Further, the response acquisition unit 207 adds the inquiry sentence output by the confirmation unit 202, the response sentence from the user U1 to the inquiry sentence, and the processed response sentence to the thread associated with the user name of the user U1.

[0058] Further, the response acquisition unit 207 may assign labels to each of the response sentences from a plurality of response models 230 by performing analysis by the first analysis unit 201 and the second analysis unit 204. Then, for each of the response sentences, the response acquisition unit 207 calculates the degree of relevance (response relevance) between the label assigned to the question sentence and the label assigned to the response sentence, and may arrange and synthesize the response sentences with high calculated response relevance so as to give priority. Further, for the response sentences whose response relevance is less than a predetermined threshold, the response acquisition unit 207 may suppress the output by the output unit 208.

[0059] When the answer acquisition unit 207 acquires an answer sentence, the selection unit 205 evaluates the answer model 230 that generated the answer sentence. The answer acquisition unit 207 assigns a label to the answer sentence generated using the second analysis unit 204. The answer acquisition unit 207 calculates an evaluation value of the answer model 230 by comparing the label assigned to the answer sentence with the label assigned to the question sentence. For example, the answer acquisition unit 207 calculates a higher evaluation value as the number of matching labels between the label assigned to the answer sentence and the label assigned to the question sentence is larger. The number of matching labels is also referred to as the degree of agreement. The selection unit 205 may update the "reliability" item in the model feature table 211 based on the calculated evaluation value. Also, when the evaluation value of the answer model 230 that generated the answer sentence is low, the answer acquisition unit 207 may select another answer model 230 and cause the analysis result input unit 206 to input the analysis result.

[0060] The output unit 208 transmits the answer sentence acquired by the answer acquisition unit 207 to the user terminal 3. When the answer acquisition unit 207 acquires answer information, the output unit 208 may transmit the acquired answer information to the user terminal 3. When the answer sentence acquired by the answer acquisition unit 207 includes information indicating the location of materials related to the answer sentence, the output unit 208 may also output the information indicating the location of the materials. The charging unit 209 calculates the usage fee of the answering system 1 based on the amount of text of the question sentence analyzed by the analysis result input unit 206 and charges the user U1.

[0061] <Processing Flow of Answering System 1> FIGS. 7 to 9 are diagrams showing an example of the processing flow of the answering device 2 according to the embodiment. Hereinafter, an example of the processing flow of the answering device 2 will be described with reference to FIGS. 7 to 9.

[0062] In step S1, a login process is performed. The login processing unit 200 executes the login process of the user U1 based on, for example, the user name and password transmitted from the user terminal 3.

[0063] In step S2, the first analysis unit 201 determines whether it has received either an input sentence of a question sentence from the user terminal 3 or an answer sentence from the answer model 230. If it has received the input sentence (YES in step S2), the process proceeds to step S3. If it has not received the input sentence (NO in step S2), the process of step S2 is repeated.

[0064] In step S3, the first analysis unit 201 generates a thread including the input sentence received in step S2. The first analysis unit 201 associates the generated thread with the user name for which the login process was performed in step S1 and stores it in the auxiliary storage unit 23. In step S4, the first analysis unit 201 identifies the language of the input sentence received in step S2.

[0065] In step S5, the first analysis unit 201 performs morphological analysis of the input sentence received in step S2 based on the language identified in step S4. The first analysis unit 201 further inputs each word included in the input sentence received in step S2 into the word model 220 selected based on the language identified in step S4 to specify the meaning of the word.

[0066] In step S6, the second analysis unit 204 assigns a label representing the meaning of each of the morphemes decomposed by the morphological analysis in step S5. The second analysis unit 204 performs clustering of each morpheme included in the input sentence so that the morphemes assigned the same label belong to the same cluster.

[0067] In step S7, when there are inquiries, the confirmation unit 202 generates an inquiry sentence. For example, when there is a word whose meaning could not be specified in step S5, the confirmation unit 202 generates an inquiry sentence about the meaning of the word. Also, when there is a word that could not be normally clustered in step S6, the confirmation unit 202 generates an inquiry sentence about the word. When an inquiry sentence is generated (YES in step S7), the process proceeds to step S2, and the processes after step S2 are executed for the generated inquiry sentence. When an inquiry sentence is not generated (NO in step S7), the process proceeds to step S8.

[0068] In step S8, the answer acquisition unit 207 determines whether the input sentence received in step S2 is an answer sentence from the answer model 230. When it is an answer sentence from the answer model 230 (YES in step S8), the process proceeds to step S9. When it is not an answer sentence from the answer model 230 (NO in step S8), the process proceeds to step S14.

[0069] In step S9, the answer acquisition unit 207 processes the answer sentence from the answer model 230 received in step S2. Examples of the processing of the answer sentence by the answer acquisition unit 207 include synthesis and summarization of answer sentences from multiple answer models 230. When the answer acquisition unit 207 completes the processing of the answer sentence, it instructs the output unit 208 to output the processed answer sentence.

[0070] In step S10, the selection unit 205 evaluates each of the answer models 230 that output the answer sentence received in step S2.

[0071] In step S11, the output unit 208 determines whether it has received an output instruction for the answer sentence from the answer acquisition unit 207 in step S9. When it has received the output instruction (YES in step S11), the process proceeds to step S12. When it has not received the output instruction (NO in step S11), the process proceeds to step S14.

[0072] In step S12, the output unit 208 outputs the response sentence processed in step S9 to the user terminal 3. When the user terminal 3 receives the response sentence processed in step S9 from the response device 2, the received response sentence is output to the output unit 36.

[0073] In step S13, the response acquisition unit 207 adds the inquiry sentence output by the confirmation unit 202, the response sentence from the user U1 to the inquiry sentence, and the processed response sentence to the thread associated with the user name of the user U1.

[0074] In step S14, the analysis result input unit 206 processes the input sentence received in step S2 into a form that makes it easier to obtain a response from the response model 230.

[0075] In step S15, the selection unit 205 calculates the degree of association between the label assigned to the input sentence received in step S2 and the "label" in the model feature table 211.

[0076] In step S16, the selection unit 205 selects one or more response models 230 based on the degree of association calculated in step S15.

[0077] In step S17, the analysis result input unit 206 inputs the input sentence processed in step S14 to the response model 230 selected in step S16.

[0078] In step S18, the response acquisition unit 207 determines whether response sentences from all the response models 230 selected in step S16 have been acquired. If acquired (YES in step S18), the process proceeds to step S4, and the processes after step S4 are executed for the acquired response sentences. If there are response models 230 that have not been acquired (NO in step S18), the process proceeds to step S17.

[0079] FIG. 10 is a diagram showing an example of a processing flow of sentence analysis processing by the first analysis unit 201 of the answering device 2 according to the embodiment. The processing flow illustrated in FIG. 10 corresponds to the processing from step S3 to S4 in FIG. 7. Hereinafter, with reference to FIG. 10, an example of a processing flow of sentence analysis processing by the first analysis unit 201 will be described.

[0080] In step S21, when the first analysis unit 201 receives an input sentence exemplified by a question sentence from the user U1 and an answer sentence from the answer model 230, the first analysis unit 201 generates a thread of the input sentence. The first analysis unit 201 stores the generated thread in the auxiliary storage unit 23 in association with the user name of the user U1.

[0081] In step S22, the first analysis unit 201 identifies the language of the question sentence for which the thread was generated in step S21. For example, the first analysis unit 201 may store information indicating the identified language in the auxiliary storage unit 23 in association with the user name of the user U1 and the thread generated in step S21.

[0082] FIG. 11 is a diagram showing an example of a processing flow of inquiry sentence generation processing by the confirmation unit 202 of the answering device 2 according to the embodiment. The processing flow illustrated in FIG. 11 corresponds to the processing in step S7 in FIG. 7. Hereinafter, with reference to FIG. 11, an example of a processing flow of inquiry sentence generation processing by the confirmation unit 202 will be described.

[0083] In step S31, the confirmation unit 202 determines whether or not the clustering by the second analysis unit 204 has been successfully performed. If the clustering has been successfully performed (YES in step S31), the process ends. If the clustering has not been successfully performed (NO in step S31), the process proceeds to step S32.

[0084] In step S32, the confirmation unit 202 generates an inquiry sentence about the meaning of the words that could not be normally clustered in step S31. In step S33, the confirmation unit 202 associates the inquiry sentence generated in step S32 with the words that could not be normally clustered in step S31 and stores them in the auxiliary storage unit 23.

[0085] FIGS. 12 and 13 are diagrams showing an example of a labeling processing flow by the first analysis unit 201 and the second analysis unit 204 of the answering device 2 according to the embodiment. The processing flow illustrated in FIGS. 12 and 13 corresponds to the processing of step S6 in FIG. 7. Hereinafter, with reference to FIG. 12, an example of a labeling processing flow by the first analysis unit 201 and the second analysis unit 204 will be described.

[0086] In step S41, the second analysis unit 204 grasps the outline of the input sentence based on the result of the morphological analysis by the first analysis unit 201.

[0087] In step S42, the first analysis unit 201 inputs each word included in the input sentence into the word model 220 to specify its meaning.

[0088] In step S43, the second analysis unit 204 uses the meaning of the words specified in step S42 to assign a label representing the meaning of these morphemes to each of the morphemes decomposed by the morphological analysis by the first analysis unit 201.

[0089] In step S44, the second analysis unit 204 determines whether another input sentence other than the current input sentence that is the target of the analysis by the first analysis unit 201 and the second analysis unit 204 is included in the thread. Examples of another input sentence include an inquiry sentence by the confirmation unit 202 and a response sentence to the inquiry sentence. If there is another input sentence (YES in step S44), the process proceeds to step S45. If there is no other input sentence (NO in step S44), the process proceeds to step S46.

[0090] In step S45, the second analysis unit 204 determines whether the content is positive or negative for the current input sentence and another input sentence. In step S46, the second analysis unit 204 determines whether the content is positive or negative for the current input sentence.

[0091] In step S47, the second analysis unit 204 assigns a label indicating whether the content is positive or negative determined in step S45 or step S46 to the current input sentence.

[0092] In step S48, the second analysis unit 204 performs clustering of each morpheme based on the labels assigned in step S43 and step S47. If the clustering is successfully completed (YES in step S49), the process proceeds to step S51. If the clustering is not successfully completed (NO in step S49), the process ends.

[0093] In step S51, the second analysis unit 204 assigns a label indicating the attributes of the user U1 who received the login process in step S1 of FIG. 7 to the current input sentence. Examples of the attributes of the user U1 include the age, gender, nationality, place of residence, etc. of the user U1. The attributes of the user U1 are stored, for example, in the auxiliary storage unit 23 in association with the user name when the user registers with the answering device 2.

[0094] In step S52, the second analysis unit 204 stores the current input sentence, the words with their meanings specified, and the assigned labels in association with each other in the auxiliary storage unit 23.

[0095] FIGS. 14 to 16 are diagrams showing an example of the processing flow of the selection process of the answer model 230 by the second analysis unit 204 and the selection unit 205 of the answering device 2. The processing of FIGS. 14 to 16 corresponds to the processing of step S16 in FIG. 9. Hereinafter, with reference to FIGS. 14 to 16, an example of the processing flow of the selection process of the answer model 230 by the second analysis unit 204 and the selection unit 205 will be described.

[0096] In step S61, the selection unit 205 determines whether it is an input to the answer model 230. If it is an input to the answer model 230 (YES in step S61), the process proceeds to step S62. If it is not an input to the answer model 230 (NO in step S61), the process proceeds to step S68.

[0097] In step S62, the selection unit 205 extracts candidates for the answer model 230 to be used in the answer based on the clustering result by the second analysis unit 204 and the assigned label.

[0098] In step S63, the selection unit 205 selects the answer model 230 to be used in the answer from the candidates extracted in step S62. If there are multiple selected answer models 230 (YES in step S64), the process proceeds to step S65. If there are not multiple selected answer models 230 (NO in step S65), the process proceeds to step S67. In step S65, the second analysis unit 204 determines the answer required by the question sentence based on the assigned label and the clustering result. Examples of the answer required by the question sentence include wanting multiple reference answers, wanting a more accurate answer, etc.

[0099] In step S66, the selection unit 205 associates the input sentence received in step S2 of FIG. 7 with the answer model 230 selected in step S63 and stores it in the auxiliary storage unit 23.

[0100] In step S67, the selection unit 205 calculates the degree of relevance between the answer model 230 selected in step S63 and the input sentence received in step S2 of FIG. 7.

[0101] In step S68, the selection unit 205 determines whether to switch the answer model 230. If it is a switch (YES in step S68), the process proceeds to step S69. If it is not a switch (NO in step S68), the process proceeds to step S61.

[0102] In step S68, the selection unit 205 determines whether to switch the answer model 230. If it is a switch (YES in step S68), the process proceeds to step S69. If it is not a switch (NO in step S68), the process proceeds to step S61.

[0103] In step S69, the selection unit 205 determines whether there is a response model 230 for which the analysis result has not been input by the analysis result input unit 206. If it exists (YES in step S69), the process proceeds to step S70. If it does not exist (NO in step S69), the process ends.

[0104] In step S70, the selection unit 205 selects, for example, according to the relevance calculated in step S67, a response model 230 for which the analysis result has not yet been input from among the response models 230 selected in step S63, and for which the analysis result is to be input by the analysis result input unit 206.

[0105] FIG. 16 is a diagram showing an example of a processing flow of the evaluation process of the response model 230 by the selection unit 205 of the response device 2 according to the embodiment. The process of FIG. 16 corresponds to the process of step S10 in FIG. 8. Hereinafter, with reference to FIG. 16, an example of a processing flow of the evaluation process of the response model 230 by the selection unit 205 will be described.

[0106] In step S81, the selection unit 205 acquires the input sentence stored in the auxiliary storage unit 23 and the label assigned to the input sentence.

[0107] In step S82, the selection unit 205 acquires the response sentence from the response model 230 stored in the auxiliary storage unit 23 and the label assigned to the response.

[0108] In step S83, the selection unit 205 calculates an evaluation value of the response model 230 that generated the response sentence acquired in step S82 based on the degree of match (also referred to as "relevance") between the label acquired in step S81 and the label acquired in step S82. If the calculated evaluation value is equal to or greater than the threshold value (YES in step S84), the process proceeds to step S85. If the calculated evaluation value is less than the threshold value (NO in step S84), the process proceeds to step S86.

[0109] In step S85, the selection unit 205 instructs the output unit 208 to output the response sentence output by the response model 230. In step S86, the selection unit 205 switches the response model 230 used for the response to another response model 230 selected in step S63 of FIG. 14. switch.

[0110] <Example> Regarding the embodiments described above, a more specific example will be described based on a specific example. In this example, an example in which a question sentence about travel is received and the answering device 2 answers the question sentence will be described.

[0111] FIG. 17 is a diagram showing an example of a user screen 301 displayed on the output unit 36 of the user terminal 3 according to the example. The user screen 301 is used for input of a question sentence by the user U1, and output of an inquiry sentence and a response sentence from the user terminal 3. The user screen 301 includes a history area 302 and a question / answer area 303. Also, on the user screen 301, a user icon 304 indicating that it is an input sentence from the user U1 is attached to the input sentence from the user U1. Further, on the user screen 301, a response device icon 305 (see FIG. 18) indicating that it is an output sentence from the response device 2 is attached to the output sentence from the response device 2. In the history area 302, a history of items that have been questioned so far is displayed. In the question / answer area 303, input of a question sentence, and display of an inquiry sentence and a response sentence from the user terminal 3 are performed. In the example of FIG. 17, a state in which a question "I want to go on a hot spring trip in January next year. Which place is recommended?" from the user U1 is input is illustrated. The input question sentence is sent from the user terminal 3 to the response device 2 via the network N1.

[0112] The first analysis unit 201 generates a thread for the received question sentence. Also, the first analysis unit 201 identifies the language of the received question sentence, for example, as described in step S4 of FIG. 7 (step S4 of FIG. 7). Further, the first analysis unit 201 performs morphological analysis on the question sentence. The following Table 1 is a table exemplifying the results of morphological analysis. In Table 1, the "written form" stores the words extracted from the question sentence by morphological analysis. The "lexical morpheme" stores the notation that groups the words in the written form regardless of the difference in inflected forms. The "reading of lexical morpheme" stores the katakana notation indicating the reading of the lexical morpheme. The "part of speech" stores information indicating the part of speech of the morpheme. The first analysis unit 201 may store the results of morphological analysis in the auxiliary storage unit 23 in association with "written form", "lexical morpheme", "reading of lexical morpheme", and "part of speech" as shown in Table 1.

Table 1

[0113] The second analysis unit 204 assigns a label to each morpheme of the question sentence. The second analysis unit 204 extracts, for example, the lexical morphemes of the morphemes whose part of speech is a noun, verb, final particle, or auxiliary symbol from the results of morphological analysis. In this embodiment, "next year", "one", "month", "hot spring", "travel", "go", "recommend", "ka", "?" are extracted. The second analysis unit 204 may, for example, instruct the first analysis unit 201 to input each of the extracted morphemes into the word model 220 and obtain the meaning of each morpheme.

[0114] Further, the second analysis unit 204 may recognize that it is "January 2024" by concatenating "next year", "one", and "month". Then, based on the meanings obtained from the word model 220 and the meanings of auxiliary symbols, the second analysis unit 204 assigns a label of "hot spring" to "hot spring", a label of "travel" to "travel", a label of "January 2024" to "next year" and "January", a label of "hope" to "go", "ta", and "i", a label of "recommend" to "o" and "susume", and a label of "interrogative form" to "desu", "ka", and "?". Then, based on these assigned labels, the second analysis unit 204 assigns labels of "travel plan search" and "travel plan proposal" to the entire question sentence.

[0115] Based on the result of the analysis by the second analysis unit 204, the selection unit 205 recognizes that the question sentence relates to travel plan search and travel plan proposal. The selection unit 205 calculates the degree of relevance between the labels assigned to the question sentence and each morpheme included in the question sentence and the labels associated with each of the answer models 230 in the model feature table 211, and selects an answer model 230 to be used for generating an answer sentence for the question sentence based on the calculated degree of relevance. When the model feature table 211 is in the state illustrated in FIG. 6, the first travel model 234 in which the two labels of "travel" and "hot spring" match among the answer models 230 is selected. When the analysis result by the second analysis unit 204 is input to the first travel model 234 by the analysis result input unit 206, an answer sentence is output from the first travel model 234. Here, when there are a large number (equal to or more than a predetermined threshold) of corresponding travel plans, the first travel model 234 outputs an answer sentence indicating that the conditions are insufficient. Here, it is assumed that an answer sentence such as "Conditions such as the desired area and budget are insufficient. There are a large number of corresponding plans." is obtained from the first travel model 234.

[0116]

[0117] ​Here, the confirmation unit 202 outputs, for example, the response sentence obtained from the first travel model 234 as an inquiry sentence for the user U1 to the user terminal 3. The user terminal 3 displays the inquiry sentence received from the response device 2 on the user screen 301. FIG. 18 is a diagram illustrating a state in which the inquiry sentence received from the response device 2 is displayed on the user screen 301. In the example of FIG. 18, an inquiry sentence "Conditions such as the desired area and budget are insufficient. There are many corresponding plans." is displayed as the inquiry sentence. By displaying the inquiry sentence from the response device 2 on the user screen 301, the response device 2 can prompt the user U1 to supplement the insufficient information in the inquiry sentence.

[0118] FIG. 19 is a diagram illustrating the user screen 301 in which a response sentence to the inquiry sentence from the response device 2 is input by the user U1. In the example of FIG. 19, a response sentence "I am thinking about staying for about two nights and three days in the vicinity of Kanto. I hope to have a hot spring with good water quality." is input. The input response sentence is sent from the user terminal 3 to the response device 2 via the network N1.

[0119] The response device 2 adds the received response sentence to the thread. Then, in the response device 2, for the received inquiry sentence, the meaning of the response sentence is grasped by the processing of, for example, the first analysis unit 201 and the second analysis unit 204. Then, in the response device 2, based on the inquiry sentence and the response sentence included in the thread, the response model 230 is selected by the selection unit 205, and a response sentence is output. FIG. 20 is a diagram illustrating a state in which the response sentence from the response device 2 is displayed on the user screen 301. In the example of FIG. 20, a response sentence "If you are looking for a hot spring with good water quality in the vicinity of Kanto, the following areas are recommended. 1. Hakone: Famous for sulfur hot springs..." is output.

[0120] In this way, a response sentence to the inquiry sentence received from the response device 2 is input by the user U1 and sent to the response device 2. Then, by repeating such an interaction, the response device 2 can output a response that meets the requirements of the user U1 with higher accuracy.

[0121] <Advantages and effects of the embodiment> When the question sentence received from the user terminal 3 contains an unclear word whose meaning cannot be specified, or when clustering cannot be performed normally, the answering device 2 according to the present embodiment outputs an inquiry sentence. Then, the answering device 2 grasps the meaning of the question sentence based on the answer sentence from the user terminal 3 for the inquiry sentence. Through such an exchange of inquiry sentences and answer sentences, the answering device 2 can grasp the meaning of the question sentence from the user U1 with higher accuracy. As a result, the answering device 2 can improve the accuracy of the answer sentence for the question sentence containing unknown words.

[0122] When the meaning of the unclear word is specified, the answering device 2 according to the present embodiment performs additional learning processing on the word model 220 using the unclear word and the meaning of the specified unclear word as teacher data. Therefore , the answering device 2 can expand the words that can be specified by the word model 220. As a result, the answering device 2 can reduce the workload of causing the user U1 to answer the inquiry sentence.

[0123] The answering device 2 according to the present embodiment associates the unclear word, the meaning of the specified unclear word, and the input sentence including the unclear word and stores them in the auxiliary storage unit 23. Therefore, even if no additional learning process for the unclear word for the word model 220 is performed, the meaning of the unclear word that could not be specified by the word model 220 can be specified by referring to the auxiliary storage unit 23. As a result, the answering device 2 can reduce the workload of causing the user U1 to answer the inquiry sentence.

[0124] When the meaning of the unclear word cannot be specified by the unclear word specifying unit 203 either, the answering device 2 according to the present embodiment may notify the user terminal 3 that answering is impossible. Therefore, the answering device 2 can suppress outputting an answer that does not conform to the intention of the question sentence.

[0125] When there is material related to the answer sentence, the answering device 2 according to this embodiment also outputs information indicating the location of the material. Therefore, in addition to the answer sentence, the answering device 2 can present to the user U1 the material that is a reference for answering the question sentence.

[0126] In this embodiment, the characteristics of each model of the answer model 230 are managed by the model characteristic table 211. Then, the answering device 2 selects an answer model 230 that inputs the analysis result by the second analysis unit 204 based on the label given to the question sentence by the second analysis unit 204 and the label associated with each model in the model characteristic table 211. The label given to the question sentence indicates the field of the question sentence, and the label associated with each model indicates the field of the teacher data of the model. Therefore, the answering device 2 can select an answer model 230 that is preferable for generating an answer sentence for the question sentence.

[0127] In this embodiment, the answering device 2 calculates the degree of relevance between the label given to the question sentence and the label associated with each model in the model characteristic table 211, and preferentially selects the answer model 230 with a high calculated degree of relevance. Therefore, the answering device 2 can select, as the answer model 230 for generating an answer sentence, an answer model 230 that is more relevant to the question sentence. As a result, the answering device 2 can improve the accuracy of the answer to the question sentence. Further, when a plurality of answer models 230 are selected for generating an answer sentence, the answering device 2 inputs the analysis results in order from the answer model 230 with a high degree of relevance. Therefore, the answering device 2 can obtain an answer sentence with higher accuracy at an earlier stage.

[0128] In this embodiment, the answering device 2 calculates the answer relevance between the question sentence and the answer sentence obtained from the answer model 230, and preferentially outputs the answer sentence with a high calculated answer relevance. Further, when the calculated answer relevance is less than a predetermined threshold value, the answering device 2 suppresses the output of the answer sentence. Therefore, the answering device 2 can present to the user U1 an answer sentence that is more relevant to the question sentence.

[0129] In this embodiment, when the answering device 2 obtains answer sentences from a plurality of answer models 230, it synthesizes the plurality of answer sentences. Therefore, the answering device 2 can list the answer sentences obtained from the plurality of answer models 230 for the user U1.

[0130] In this embodiment, the answering device 2 calculates the usage fee of the answering system 1 based on the amount of text in the question sentence and charges the user U1. Therefore, the answering device 2 can bill the user U1 according to the usage mode (usage amount) of the answering system 1 by the user U1.

[0131] <Modification Example> In the embodiment described above, the process of inputting the analysis result into the answer model 230 by the analysis result input unit 206 and the process of obtaining the answer sentence from the answer model 230 by the answer acquisition unit 207 are performed continuously. However, the process of inputting the analysis result (an example of "the first analysis result") of the question sentence (an example of "the first question sentence") by the analysis result input unit 206 into the answer model 230 and the answer sentence (an example of "the first answer information") from the answer model 230 for the analysis result of the question sentence (an example of "the second question sentence") that has already been input into the answer model 230 may be executed in parallel. By executing in parallel, the answering device 2 can shorten the time until the generation of the final answer sentence for the question sentence from the user U1. The parallel execution at this time is also referred to as multitasking, which means that the input process and the acquisition process are in a state where they are processed simultaneously even for a moment, and it does not mean that the two processes are started simultaneously, but they may be started simultaneously.

[0132] In the embodiment described above, the answering device 2 and the user terminal 3 are separate devices, but the answering device 2 and the user terminal 3 may be realized by a single information processing device.

[0133] The above-described embodiments and modification examples can be combined with each other.

[0134] <Computer-readable Recording Medium> An information processing program for causing a computer or other machine or device (hereinafter referred to as a computer or the like) to implement any of the above functions can be recorded on a recording medium readable by the computer or the like. Then, by causing the computer or the like to read and execute the program of this recording medium, the function can be provided.

[0135] Here, the recording medium readable by a computer or the like refers to a recording medium that accumulates information such as data and programs by electrical, magnetic, optical, mechanical, or chemical actions and can be read by a computer or the like. Among such recording media, removable ones from a computer or the like include, for example, flexible disks, magneto-optical disks, Compact Disc Read Only Memory (CD-ROM), Compact Disc-Recordable (CD-R), Compact Disc-ReWriterable (CD-RW), Digital Versatile Disc (DVD), Blu-ray Disc (BD), Digital Audio Tape (DAT), 8mm tape, flash memory, external hard disk drives, Solid State Drive (SSD), etc. Also, as recording media fixed to a computer or the like, there are built-in hard disk drives, SSDs, ROMs, etc.

[0136] The above learning refers to a method in which data patterns are input to a computer based on data called teacher data, which is a set of information and correct judgments, and the computer finds new features and correlations based on these patterns, or as an example, a method in which the computer finds commonalities and feature data of the input data. In addition, learning is, for example, a method in which a computer constructs a neural network to analyze, extract, and output local or characteristic data, which is called deep learning or the like. Of course, this learning includes preprocessing composed of processes such as supplementing insufficient data, eliminating unnecessary data, and unifying data formats, and also includes postprocessing for storing the learned results so that they can be used in other learning.

[0137] The "parts", "means", "devices", and "systems" used in the above-described embodiments and claims do not simply mean physical means, but also include cases where the functions they have are realized by software or software services. In addition, the functions of a single "part", "means", "device", or "system" are not only constituted by a single physical means, software, software module, or device, but may also be realized by a plurality of physical means, software, software modules, devices, or combinations thereof.

[0138] The terms used in the above-described embodiments and claims should be construed as non-limiting terms. For example, the term "including" should be construed as "not limited to those described as including". The term "containing" should be construed as "not limited to those described as containing". The term "comprising" should be construed as "not limited to those described as comprising". The term "having" should be construed as "not limited to those described as having". The term "equipped with" should be construed as "not limited to those described as equipped with".

[0139] <Appendix 1> A storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs answer information for the question when the question is input; and a processor connected to the storage unit, wherein the processor performs a first specifying process of specifying the meaning of each word included in a question sentence input by a user using the first learning model, an analyzing process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, an input process of inputting the analysis result of the question sentence by the analyzing process into the second learning model, an acquisition process of acquiring first answer information for the question sentence from the second learning model, and an output process of outputting the first answer information, and the analyzing process outputs inquiry information for inquiring about the meaning of the word when at least one of the cases where the meaning cannot be specified in the first specifying process or the meaning of the word can be specified but the first answer information cannot be output by the output process based on the meaning of the word, receives second answer information for the inquiry information, and when the meaning of the word that could not be specified by the first specifying process is specified based on the second answer information, analyzes the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second answer information. An information processing apparatus further comprising: <Appendix 2> The information processing apparatus according to Appendix 1, wherein the processor further executes a learning process of training the first learning model based on the word specified based on the second answer information and the meaning of the word specified based on the second answer information. <Appendix 3> The information processing apparatus according to Appendix 1, wherein the processor further executes a storage process of storing the second answer information in the storage unit in association with the question sentence and the word for which the meaning could not be specified. <Appendix 4> The information processing apparatus according to Appendix 1, wherein specifying based on the second answer information further includes notifying that an answer is not available when the meaning of the word for which the meaning could not be specified cannot be specified based on the second answer information. <Supplementary Note 5> The information processing apparatus according to Supplementary Note 1, wherein the output processing further includes a process of including location information indicating the location of materials related to the first response information in the first response information. <Supplementary Note 6> The processor executes the input processing of inputting the first analysis result for the first question sentence into the second learning model, and the acquisition processing of acquiring the first response information for the second question sentence whose analysis result has already been input into the second learning model from the second learning model. The information processing apparatus according to any one of Supplementary Notes 1 to 5. <Supplementary Note 7> A first learning model that outputs text data indicating the meaning of a word when the word is input and a second learning model that outputs response information for a question when the question is input, and a computer including a storage unit that stores the first learning model and the second learning model, performs a first specifying process of specifying the meaning of each word included in a question sentence input by a user using the first learning model, an analysis process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, an input process of inputting the analysis result of the question sentence by the analysis process into the second learning model, an acquisition process of acquiring the first response information for the question sentence from the second learning model, and an output process of outputting the first response information. The analysis process outputs inquiry information for inquiring about the meaning of the word when at least one of the cases where the meaning cannot be specified in the first specifying process or the meaning of the word can be specified but the first response information cannot be output by the output process based on the meaning of the word occurs, receives the second response information for the inquiry information, and when the meaning of the word that could not be specified in the first specifying process is specified based on the second response information, analyzes the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second response information. The information processing method further includes the above steps. <Supplementary Note 8> A computer comprising a storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs answer information for the question when the question is input, performs a first specifying process of specifying the meaning of each word included in the question sentence input by the user using the first learning model, an analysis process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, an input process of inputting the analysis result of the question sentence by the analysis process into the second learning model, an acquisition process of acquiring first answer information for the question sentence from the second learning model, and an output process of outputting the first answer information. The analysis process outputs inquiry information for inquiring about the meaning of the word when at least one of the following cases occurs: when the meaning cannot be specified in the first specifying process, or when the meaning of the word can be specified but the first answer information cannot be output by the output process based on the meaning of the word. The second answer information for the inquiry information is received, and the first specifying process specifies the meaning of the word for which the meaning could not be specified based on the second answer information. Based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second answer information, the analysis process further includes analyzing the meaning of the question sentence. Information processing program.

[0140] <Appendix 11> When the analysis result of the article is input, a plurality of learning models that generate a response sentence for the article, and feature information indicating features related to the response accuracy for the analysis result for each of the plurality of learning models, a storage unit that stores the above, and a processor connected to the storage unit, and when the processor receives the analysis result of the question sentence input by the user, it performs an input process of inputting the analysis result to one or more selected learning models selected from the plurality of learning models, and an output process of outputting a response sentence from the selected learning model to which the analysis result has been input, and the input process further includes a selection process of selecting one or more of the selected learning models for inputting the analysis result from the plurality of learning models with reference to the feature information. An information processing apparatus. <Appendix 12> The feature information includes first field information indicating the field to which the learning model corresponds, the analysis result includes second field information indicating the field of the question sentence, the selection process calculates a field relevance indicating the relevance between the first field information and the second field information, and includes a process of preferentially selecting the learning model with the calculated high field relevance as the selected learning model. The information processing apparatus according to Appendix 11. <Appendix 13> When a plurality of the selected learning models are selected in the input process, the The analysis result is input to the selected learning model in the order according to the height of the field relevance. The information processing apparatus according to Appendix 12. <Appendix 14> The output process further includes a process of calculating a response relevance indicating the relevance between the response sentence and the second field information, and preferentially outputting the response sentence with the high response relevance. The information processing apparatus according to Appendix 12. <Appendix 15> The output process further includes a process of calculating a response relevance indicating the relevance between the response sentence and the second field information, and suppressing the output of the response sentence with the response relevance lower than the threshold value. The information processing apparatus according to Appendix 12. <Appendix 16> The output process includes a process of synthesizing the response sentences from the plurality of selected learning models when a plurality of the selected learning models are selected in the selection process, and the information processing apparatus according to Supplementary Note 11. <Supplementary Note 17> The analysis result includes text amount information related to the text amount of the sentence, and the processor further executes a calculation process of calculating a usage fee based on the text amount information, and the information processing apparatus according to any one of Supplementary Notes 11 to 16. <Supplementary Note 18> A computer including a storage unit that stores a plurality of learning models that generate response sentences for the sentence when an analysis result of the sentence is input, and feature information indicating features related to the response accuracy for the analysis result for each of the plurality of learning models. When receiving an analysis result of a question sentence input by a user, the computer executes an input process of inputting the analysis result to one or a plurality of selected learning models selected from the plurality of learning models, and an output process of outputting a response sentence from the selected learning model to which the analysis result is input. The input process further includes a selection process of selecting one or a plurality of the selected learning models for inputting the analysis result from the plurality of learning models with reference to the feature information. <Supplementary Note 19> When a computer including a storage unit that stores a plurality of learning models that generate response sentences for the sentence when an analysis result of the sentence is input, and feature information indicating features related to the response accuracy for the analysis result for each of the plurality of learning models receives an analysis result of a question sentence input by a user, the computer is caused to execute an input process of inputting the analysis result to one or a plurality of selected learning models selected from the plurality of learning models, and an output process of outputting a response sentence from the selected learning model to which the analysis result is input. The input process further includes a selection process of selecting one or a plurality of the selected learning models for inputting the analysis result from the plurality of learning models with reference to the feature information. Information processing program.

Explanation of Signs

[0141] 1 ··· Response system 2. Answer device 3. User terminal 21. CPU 22. Main memory unit 23. Auxiliary memory unit 24. Communication unit 31. CPU 32. Main memory unit 33. Auxiliary memory unit 34. Communication unit 35. Input unit 36. Output unit 200. Login processing unit 201. First analysis unit 202. Confirmation unit 203. Unknown word identification unit 204. Second analysis unit 205. Selection unit 206. Analysis result input unit 207. Answer acquisition unit 208. Output unit 209. Billing unit 210. Management database 211. Model feature table 220. Word model 230. Answer model 231. First maintenance model 232. Second maintenance model 233. Third maintenance model 234. First travel model 235. Second travel model 301. User screen 302. History area 303. Question and answer area 304. User icon 305. Answer device icon B2. Connection bus B3. Connection bus N1. Network U1. User

Claims

1. A storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs answer information for the question when the question is input; A processor connected to the storage unit; The processor: A first specifying process of specifying the meaning of each word included in the question sentence input by the user using the first learning model; An analysis process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process; An input process of inputting the analysis result of the question sentence by the analysis process into the second learning model; An acquisition process of acquiring first answer information for the question sentence from the second learning model; An output process of outputting the first answer information; The analysis process: When the first specifying process cannot specify, or when the meaning of the word can be specified but the first answer information cannot be output by the output process based on the meaning of the word, output inquiry information for inquiring about the meaning of the word; Receiving second answer information for the inquiry information, and when the first specifying process specifies the meaning of the word that could not be specified based on the second answer information, analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second answer information; An information processing apparatus.

2. The processor: Further executes a learning process of learning the first learning model based on the word specified based on the second answer information and the meaning of the word specified based on the second answer information; The information processing apparatus according to claim 1.

3. The processor: Further executes a storage process of storing the second answer information in the storage unit in association with the question sentence and the word whose meaning could not be specified; The information processing apparatus according to claim 1.

4. Specifying based on the second answer information further includes notifying that the answer is unavailable when the meaning of the word whose meaning could not be specified cannot be specified based on the second answer information; The information processing apparatus according to claim 1.

5. The output process further includes a process of including location information indicating the location of materials related to the first answer information in the first answer information; The information processing apparatus according to claim 1.

6. The processor executes the input process of inputting a first analysis result for a first question sentence into the second learning model, and the acquisition process of acquiring the first answer information for a second question sentence whose analysis result has already been input into the second learning model from the second learning model. The information processing apparatus according to any one of claims 1 to 5.

7. A computer including a storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs answer information for a question when the question is input, performs a first specifying process of specifying the meaning of each word included in a question sentence input by a user using the first learning model, an analysis process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, an input process of inputting an analysis result of the question sentence by the analysis process into the second learning model, an acquisition process of acquiring first answer information for the question sentence from the second learning model, and an output process of outputting the first answer information, wherein the analysis process outputs inquiry information for inquiring about the meaning of the word when the meaning cannot be specified in the first specifying process, or when the meaning of the word can be specified but the first answer information cannot be output by the output process based on the meaning of the word, receives second answer information for the inquiry information, and when the first specifying process specifies the meaning of the word that could not be specified based on the second answer information, analyzes the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second answer information. An information processing method.

8. In a computer including a storage unit that stores a first learning model that outputs text data indicating the meaning of a word when the word is input, and a second learning model that outputs answer information for a question when the question is input, a first specifying process of specifying the meaning of each word included in a question sentence input by a user using the first learning model, an analysis process of analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process, An input process of inputting the analysis result of the question sentence by the analysis process into the second learning model; An acquisition process of acquiring first answer information for the question sentence from the second learning model; An output process of outputting the first answer information, and causing the above to be executed; The analysis process is as follows: When the meaning cannot be specified in the first specifying process, or when the meaning of the word can be specified but the first answer information cannot be output by the output process based on the meaning of the word, output inquiry information for inquiring about the meaning of the word; Receiving second answer information for the inquiry information, causing the meaning of the word for which the meaning could not be specified by the first specifying process to be specified based on the second answer information, and analyzing the meaning of the question sentence based on the meaning of the word specified in the first specifying process and the meaning of the word specified based on the second answer information; further including this; An information processing program.

Images