Information processing program, server and information processing system

Abstract

To provide an information processing program, a server and an information processing system for performing processing to convert at least a part of an answer sentence from an interactive AI into "hiragana" (the cursive form of the Japanese syllabary).SOLUTION: An information processing program comprises: acquiring and inputting an input sentence from a user terminal to a predetermined language model to output an answer sentence; analyzing the answer sentence by predetermined analysis means to acquire reading information; and converting at least a part of "kanji" (Chinese character) of the answer sentence into "hiragana" or "katakana" (the square form of the Japanese syllabary) based upon the reading information according to a predetermined conversion rule, wherein a previously registered word which is included in the answer sentence is not converted and previously set processing on the registered word is executed.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] At least one embodiment of the present invention relates to an information processing program, a server, and an information processing system for converting at least a part of a response sentence from a dialogue type AI into hiragana. [Background technology]

[0002] In recent years, an interactive service called a chatbot has been provided. A chatbot can generate and present answers to users' questions, providing an environment in which the user feels as if they are having a dialogue.

[0003] For example, Patent Document 1 discloses a technique for controlling a chatbot that can maintain a dialogue body having the persona of the chatbot even while continuing a dialogue with a user. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2022-180282 A Summary of the Invention [Problem to be solved by the invention]

[0005] Meanwhile, services using conversational AI such as chatbots are increasingly using pre-trained language models to generate answer sentences. In particular, large-scale language models trained on huge amounts of data are capable of generating natural and natural answer sentences and are beginning to be used in a variety of fields.

[0006] Although such conversational AI services can generate answers to input sentences, the answers are basically generated as sentences using kanji, so depending on the user's age, there is a risk that the answers may include kanji whose pronunciation is unknown.

[0007] In addition, when a user gives an anthropomorphized character to a conversational AI that is a conversational partner, in order to obtain a response sentence that expresses the personality of that character, it is necessary to either devise an output that expresses the personality in the prompt section, or to train the model that generates the sentences itself in advance with content that can express the personality.

[0008] An object of at least one embodiment of the present invention is to provide an information processing program, a server, and an information processing system for converting at least a portion of a response sentence from a dialogue-type AI into hiragana. [Means for solving the problem]

[0009] From a non-limiting perspective, an information processing program according to one embodiment of the present invention is an information processing program for causing a server to realize a process of generating an answer to an input sentence from a user terminal using an interactive AI, and causes the server to realize an input sentence acquisition function of acquiring the input sentence from the user terminal, an answer sentence generation function of inputting the input sentence into a predetermined language model to output an answer sentence, a reading information acquisition function of analyzing the answer sentence using a predetermined analysis means to acquire reading information, and a conversion function of converting at least a portion of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with predetermined conversion rules, wherein the conversion function is characterized in that when a pre-registered word is included in the answer sentence, the registered word is not subject to conversion and a pre-set process is executed for the registered word.

[0010] From a non-limiting perspective, a server according to one embodiment of the present invention is a server for executing a process of generating an answer to an input sentence from a user terminal using an interactive AI, and includes an input sentence acquisition unit that acquires the input sentence from the user terminal, an answer sentence generation unit that inputs the input sentence into a predetermined language model to output an answer sentence, a reading information acquisition unit that analyzes the answer sentence using a predetermined analysis means to acquire reading information, and a conversion unit that converts at least a portion of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with predetermined conversion rules, wherein when a pre-registered word is included in the answer sentence, the conversion unit executes a pre-set process for the registered word without subjecting the registered word to conversion.

[0011] From a non-limiting perspective, an information processing system according to one embodiment of the present invention is an information processing system comprising a communication network, a server, and a user terminal, and for performing processing to generate an answer to an input sentence from the user terminal using an interactive AI, and includes an input sentence acquisition means for acquiring the input sentence from the user terminal, an answer sentence generation means for inputting the input sentence into a predetermined language model and outputting an answer sentence, a reading information acquisition means for analyzing the answer sentence using a predetermined analysis means to acquire reading information, and a conversion means for converting at least a portion of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with predetermined conversion rules, wherein when a pre-registered word is included in the answer sentence, the conversion means executes a pre-set process on the registered word without subjecting the registered word to conversion. Effect of the Invention

[0012] Each embodiment of the present application addresses one or more of the deficiencies. [Brief description of the drawings]

[0013] [Figure 1] 1 is a block diagram showing an example of a configuration of an information processing system corresponding to at least one of the embodiments of the present invention. [Diagram 2] FIG. 2 is a block diagram showing a configuration of a server corresponding to at least one of the embodiments of the present invention. [Diagram 3] 1 is a flowchart illustrating an example of information processing corresponding to at least one of the embodiments of the present invention. [Figure 4] 10 is a flowchart illustrating an example of an operation on a server side in information processing corresponding to at least one of the embodiments of the present invention. [Diagram 5] 10 is a flowchart illustrating an example of an operation on a terminal side in information processing corresponding to at least one of the embodiments of the present invention. [Figure 6] FIG. 2 is a block diagram showing a configuration of a server corresponding to at least one of the embodiments of the present invention. [Figure 7] 1 is a flowchart illustrating an example of information processing corresponding to at least one of the embodiments of the present invention. [Figure 8] FIG. 2 is a block diagram showing a configuration of a server corresponding to at least one of the embodiments of the present invention. [Figure 9] 1 is a flowchart illustrating an example of information processing corresponding to at least one of the embodiments of the present invention. [Figure 10] 10 is a flowchart illustrating an example of an operation on a server side in information processing corresponding to at least one of the embodiments of the present invention. [Figure 11] 10 is a flowchart illustrating an example of an operation on a terminal side in information processing corresponding to at least one of the embodiments of the present invention. [Figure 12] FIG. 11 is an explanatory diagram for explaining an example of a conversion process of an answer sentence in information processing corresponding to at least one of the embodiments of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, examples of embodiments of the present invention will be described with reference to the drawings. Note that the various components in the examples of the embodiments described below can be appropriately combined to the extent that no inconsistencies or the like arise. In addition, the contents described as an example of one embodiment may be omitted in other embodiments. In addition, the contents of operations and processes that are not related to the characteristic parts of each embodiment may be omitted. Furthermore, the order of various processes that make up the various flows described below is random to the extent that no inconsistencies or the like arise in the process contents.

[0015] [First embodiment] Fig. 1 is a block diagram showing an example of a configuration of an information processing system 100 corresponding to at least one of the embodiments of the present invention. As shown in Fig. 1, the information processing system 100 includes a server 10 and user terminals 20, 201 to 20N (N is an arbitrary integer) used by users of the information processing system 100. Note that the configuration of the information processing system 100 is not limited to this, and it may be configured such that a single user terminal is used by multiple users, or may be configured to include multiple servers.

[0016] The server 10 and the multiple user terminals 20, 201 to 20N are each connected to a communication network 30 such as the Internet. Although not shown, the multiple user terminals 20, 201 to 20N are connected to the communication network 30 by performing data communication with a base station managed by a communication carrier via a wireless communication line.

[0017] The information processing system 100 includes a server 10 and a plurality of user terminals 20, 201 to 20N, thereby realizing various functions for executing various processes.

[0018] The server 10 is managed by an administrator of the information processing system 100, and has various functions for providing information on various processes to a plurality of user terminals 20, 201 to 20N. In this example, the server 10 is configured by an information processing system such as a WWW server, and includes a storage medium for storing various information. The configuration of the server 10 is not particularly limited as long as it includes general components for performing various processes as a computer, such as a control unit and a communication unit. An example of the hardware configuration of the server 10 will be briefly described below.

[0019] As shown in FIG. 1, the server 10 includes at least a CPU (Central Processing Unit) 101, a memory 102, and a storage device 103.

[0020] The CPU 101 is a central processing unit that performs various calculations and control. If the server 10 is equipped with a GPU (Graphics Processing Unit), some of the various calculations and control may be performed by the GPU. The server 10 executes various information processing by the CPU 101 using data appropriately read into the memory 102, and stores the obtained processing results in the storage device 103 as necessary.

[0021] The storage device 103 functions as a storage medium for storing various information. The configuration of the storage device 103 is not particularly limited, but it is preferable that the storage device 103 is configured to be capable of storing all of the various types of necessary information from the viewpoint of reducing the processing load on each of the multiple user terminals 20, 201 to 20N. Examples of such a configuration include an HDD and an SSD. However, the storage unit that stores the various types of information only needs to have a storage area accessible by the server 10, and may be configured to have a dedicated storage area outside the server 10, for example.

[0022] 2 is a block diagram showing a configuration of a server 10A which is an example of the server 10 corresponding to at least one of the embodiments of the present invention. As shown in FIG. 2, the server 10A includes at least an input sentence acquisition unit 11, an answer sentence generation unit 12, a reading information acquisition unit 13, and a conversion unit 14.

[0023] The input text acquisition unit 11 has a function of acquiring input text from a user terminal. The input text may be any text that satisfies a predetermined regulation in the dialogue AI service provided by the information processing system 100 of this example. One example of the predetermined regulation is a regulation on the number of characters or a content regulation. The input text entered by the user for the dialogue AI service through the user terminal is acquired by the input text acquisition unit 11.

[0024] The answer sentence generating unit 12 has a function of inputting an input sentence into a predetermined language model and outputting an answer sentence. Here, the language model used in this example refers to a model that generates a sentence suitable as an answer to the input sentence based on a probability distribution regarding a word string. The language model may be any type that can generate an answer sentence suitable for the service by the dialogue type AI provided by the information processing system 100 of this example. In addition, the language model may be a large-scale language model. The sentence generated by the language model is obtained as an answer sentence to the input sentence. Note that, although the description is given using the expression "language model" in this example, any model mechanism can be adopted as long as it can generate an answer sentence to the input sentence.

[0025] The reading information acquisition unit 13 has a function of analyzing the answer sentence by a predetermined analysis means to acquire reading information. Here, the predetermined analysis means may be any means capable of acquiring reading information from the answer sentence. One example of an analysis means is to break down the answer sentence into parts of speech by morphological analysis, and then acquire reading information by a dictionary-based method or a machine learning-based method. The reading information acquisition unit 13 acquires reading information of the kanji part of the answer sentence.

[0026] The conversion unit 14 has a function of converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information according to a predetermined conversion rule. The conversion unit 14 also has a function of, when a registered word registered in advance is included in the answer sentence, executing a preset process for the registered word without converting the registered word. Since the reading information acquisition unit 13 acquires the reading information for all the kanji included in the answer sentence, the conversion unit 14 can convert the entire answer sentence into hiragana or katakana. However, the conversion unit 14 executes a preset process for the registered word without converting the registered word registered in advance. The registered words can be used in various ways, and one example is to register reserved words and non-conversion words. A reserved word is a word that is determined to be replaced with a character string registered in advance when the word is included in the answer sentence. For example, a reserved word that replaces "tsukimizato" with the character string "yamanashi", a reserved word that is intentionally replaced with katakana as an exception to the process of converting to hiragana, and a reserved word that is replaced with an English word are possible. Moreover, the non-substitution word is a word for which it is determined that the kanji is used as it is without converting it to hiragana or katakana. For example, it is possible to set a person's name or a proper noun as a non-substitution word, or to set a non-substitution word according to the literacy level of the assumed user. In addition, the registered word to be applied may be changed appropriately depending on the situation. For example, it is possible to specify the literacy level of the assumed user, and to apply different non-substitution words for each literacy level. In this way, the conversion unit 14 executes the process set for each registered word, and converts words other than the registered word into hiragana or katakana, so that the answer sentence generated by the language model can be converted into one that includes an appropriate amount of kanji or a pre-registered character string.

[0027] Each of the multiple user terminals 20, 201 to 20N is managed by a user, and is configured, for example, by a personal computer, a mobile phone terminal, or a PDA (Personal Digital Assistant). Note that the configuration of the user terminal that may be included in the information processing system 100 is not limited to the above-mentioned example.

[0028] Each of the plurality of user terminals 20, 201 to 20N is connected to the communication network 30 and includes hardware and software for communicating with the server 10 to execute various processes.

[0029] Next, the operation of the information processing system 100 of this embodiment will be described.

[0030] 3 is a flowchart showing an example of information processing corresponding to at least one of the embodiments of the present invention. Hereinafter, an example will be described in which the server 10A and the user terminal 20 execute information processing.

[0031] The information processing is started by the user terminal 20 transmitting an input sentence to the server 10A (step S11). The server 10A acquires the input sentence from the user terminal 20 (step S12). Next, the server 10A generates an answer sentence using a language model (step S13). Next, the server 10A acquires reading information for the answer sentence (step S14). Next, the server 10A executes processing set for each registered word for registered words included in the answer sentence, and converts other kanji characters to hiragana (step S15). Next, the server 10A outputs the converted answer sentence (step S16). Then, the user terminal 20 receives the converted answer sentence (step S17) and ends the information processing.

[0032] 4 is a flowchart showing an example of the operation on the server side in information processing corresponding to at least one of the embodiments of the present invention. Here, the operation of the server 10A in the information processing system 100 will be described again.

[0033] The information processing starts by acquiring an input sentence in the server 10A (step S101). Next, the server 10A generates an answer sentence using a language model (step S102). Next, the server 10A acquires reading information for the answer sentence (step S103). Next, the server 10A executes processing set for each registered word for registered words included in the answer sentence, and converts other kanji characters to hiragana (step S104). The server 10A then outputs the converted answer sentence (step S105), and ends the information processing.

[0034] 5 is a flowchart showing an example of the operation of the terminal side in information processing corresponding to at least one of the embodiments of the present invention. The following describes an example in which the user terminal 20 executes information processing by itself. Note that the configuration of the terminal 20 has the same functions as the configuration of the server 10A except for receiving various information from the server 10A, so description will be omitted to avoid duplication.

[0035] The information processing starts by acquiring an input sentence in the user terminal 20 (step S201). Next, the user terminal 20 generates an answer sentence using a language model (step S202). Next, the user terminal 20 acquires reading information for the answer sentence (step S203). Next, the user terminal 20 executes processing set for each registered word for registered words included in the answer sentence, and converts other kanji characters into hiragana (step S204). Then, the user terminal 20 outputs the converted answer sentence (step S205), and ends the information processing.

[0036] As described above, as one aspect of the first embodiment, the server 10A is equipped with an input sentence acquisition unit 11, an answer sentence generation unit 12, a reading information acquisition unit 13, and a conversion unit 14, acquires an input sentence from a user terminal, inputs the input sentence into a predetermined language model to output an answer sentence, analyzes the answer sentence using a predetermined analysis means to acquire reading information, and converts at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with predetermined conversion rules. In the conversion, if a pre-registered word is included in the answer sentence, the registered word is not subject to conversion and a pre-set process is executed for the registered word. Therefore, by executing the process set for each registered word and converting words other than the registered word into hiragana or katakana, the answer sentence generated by the language model can be converted into one that appropriately includes kanji and pre-registered character strings.

[0037] In other words, when the answer sentence generated by the language model is converted into hiragana or katakana after the fact, registered words such as non-substitution words, which are words that use kanji characters as they are, and reserved words that are converted into predetermined character strings, are extracted and processing is performed for each registered word, making it possible to convert the answer sentence into the desired state.

[0038] [Second embodiment] FIG. 6 is a block diagram showing a configuration of a server 10B which is an example of the server 10 corresponding to at least one of the embodiments of the present invention. In this example, the server 10B includes at least an input sentence acquisition unit 11, an answer sentence generation unit 12, a reading information acquisition unit 13, and a conversion unit 14B. In this second embodiment, a case where an information processing system is applied to a video game is described. Specifically, the information processing system is used to realize a conversation between a user and a non-player character (hereinafter, referred to as an NPC) in a video game. Note that, regarding the configurations with the same reference numerals as those in the first embodiment, the description of the configurations that function in the same way as in the first embodiment is omitted.

[0039] The conversion unit 14B has a function of identifying an NPC with which the user will converse and / or identifying a registered word to be applied by referring to the game situation when the input sentence is acquired from the user terminal, and if the registered word to be applied is included in the answer sentence, executing a preset process for the registered word without making the registered word a conversion target, and converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information according to a predetermined conversion rule for other words. Here, the game situation refers to all information that represents the situation of the video game related to the application of the registered word. Examples of information that represent the game situation include the user's playing time, the progress of the story, the level of the user character, information on which character the user uses, and the situation of the relationship with each NPC.

[0040] As a premise for the processing in the conversion unit 14B, a registered word to be applied is registered for each NPC and / or for each game situation. That is, it is possible to set the registered word to be applied differently depending on the NPC, and also to set the registered word to be applied differently depending on the game situation. This makes it possible to express the growth of an NPC according to the progress of the game, for example, by setting the registered words so that the answer sentence for a young NPC is all converted to hiragana, and as the game situation progresses and the NPC grows, the answer sentence gradually uses kanji. Also, in a situation where the NPC is a foreign NPC and language comprehension is uncertain, an answer sentence is generated using katakana or a unique character string (which may be an unreadable character string), but in order to express that the NPC's language comprehension has progressed as the game situation progresses, it is possible to make the answer sentence gradually use more hiragana and kanji.

[0041] 7 is a flowchart showing an example of information processing corresponding to at least one of the embodiments of the present invention. The following describes an example of the operations of the server 10B and the user terminal 20. Note that the flowcharts showing the operations of the server 10B and the user terminal 20 are omitted to avoid redundant description.

[0042] The information processing is started by the user terminal 20 transmitting an input sentence to the server 10B (step S21). The server 10B acquires the input sentence from the user terminal 20 (step S22). Next, the server 10B generates an answer sentence using a language model (step S23). Next, the server 10B acquires reading information for the answer sentence (step S24). Next, the server 10B identifies an NPC with which the user will converse, and / or identifies a registered word to be applied by referring to the game situation when the input sentence was acquired from the user terminal, and executes a process set for each registered word for the registered word included in the answer sentence, and converts the other kanji characters to hiragana (step S25). Next, the server 10B outputs the converted answer sentence (step S26). Then, the user terminal 20 receives the converted answer sentence (step S27), and ends the information processing.

[0043] As described above, as one aspect of the second embodiment, the server 10B includes an input sentence acquisition unit 11, an answer sentence generation unit 12, a reading information acquisition unit 13, and a conversion unit 14B, among which the conversion unit 14B identifies the NPC with which the user will converse and / or identifies a registered word to be applied by referring to the game situation at the time the input sentence is acquired from the user terminal, and if the registered word to be applied is included in the answer sentence, the registered word is not subject to conversion and a preset process is executed for the registered word, and for other words, at least a part of the kanji in the answer sentence is converted to hiragana or katakana based on the reading information in accordance with a predetermined conversion rule. Therefore, it is possible to change the registered word to be applied and the conversion content of the answer sentence depending on who the NPC with which the user is conversing is and what the game situation is.

[0044] In other words, by setting different registered words for each NPC and each game situation, it is possible to reflect the personality of the NPC or the game situation, thereby creating a sense of immersion in the video game without the player feeling that the answers are coming from a language model.

[0045] [Third embodiment] 8 is a block diagram showing a configuration of a server 10Z which is an example of the server 10 corresponding to at least one of the embodiments of the present invention. In this example, the server 10Z includes at least an input sentence acquisition unit 11Z, an answer sentence generation unit 12Z, a reading information acquisition unit 13Z, and a conversion unit 14Z.

[0046] The input text acquisition unit 11Z has a function of acquiring an input text from a user terminal. The input text may be any text that satisfies a predetermined regulation in the dialogue AI service provided by the information processing system 100 of this example. One example of the predetermined regulation is a regulation of the number of characters or a content regulation. The input text entered by the user for the dialogue AI service through the user terminal is acquired by this input text acquisition unit 11Z.

[0047] The answer sentence generating unit 12Z has a function of inputting an input sentence into a predetermined language model and outputting an answer sentence. Here, the language model used in this example refers to a model that generates a sentence suitable as an answer to the input sentence based on a probability distribution regarding a word string. The language model may be any type that can generate an answer sentence suitable for the service by the dialogue type AI provided by the information processing system 100 of this example. In addition, the language model may be a large-scale language model. The sentence generated by the language model is obtained as an answer sentence to the input sentence. Note that, although the description is given using the expression "language model" in this example, any model mechanism can be adopted as long as it can generate an answer sentence to the input sentence.

[0048] The reading information acquisition unit 13Z has a function of analyzing the answer sentence by a predetermined analysis means to acquire reading information. Here, the predetermined analysis means may be any means capable of acquiring reading information from the answer sentence. One example of the analysis means is to break down the answer sentence into parts of speech by morphological analysis, and then acquire reading information by a dictionary-based method or a machine learning-based method. The reading information acquisition unit 13Z acquires reading information of the kanji part of the answer sentence.

[0049] The conversion unit 14Z has a function of converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information according to a predetermined conversion rule. In addition, the conversion unit 14 has a function of, when a registered word registered in advance is included in the answer sentence, executing a preset process on the registered word without converting the registered word. Since the reading information acquisition unit 13 acquires the reading information of all the kanji included in the answer sentence, the conversion unit 14Z can convert the entire answer sentence into hiragana or katakana, but the conversion unit 14Z executes a preset process on the registered word without converting the registered word registered in advance. The registered words can be used in various ways, and one example is to register reserved words and non-conversion words. A reserved word is a word that is determined to be replaced with a character string registered in advance when the word is included in the answer sentence. For example, reserved words that replace "Tsukimizato" with the character string "Yama-nashi", reserved words that are intentionally replaced with katakana as an exception to the process of converting to hiragana, reserved words that are replaced with English words, etc. are conceivable. In addition, a non-replacement word is a word that is determined to use kanji as it is without converting to hiragana or katakana. For example, it is conceivable to set a person's name or a proper noun as a non-replacement word according to the literacy level of an assumed user. In addition, the registered word to be applied may be changed appropriately depending on the situation. For example, it is conceivable to make it possible to specify the literacy level of an assumed user, and to apply different non-replacement words for each literacy level. In this way, the conversion unit 14Z executes the process set for each registered word, and converts words other than the registered word into hiragana or katakana, so that the answer sentence generated by the language model can be converted to one that appropriately contains kanji and pre-registered character strings.

[0050] The information processing system can also be used to realize a conversation between a user and a non-player character (hereinafter, referred to as an NPC) in a video game. In this case, the conversion unit 14Z identifies an NPC with which the user will talk and / or identifies a registered word to be applied by referring to the game situation when an input sentence is acquired from the user terminal, and when the registered word to be applied is included in an answer sentence, the conversion unit 14Z has a function of performing a preset process on the registered word without making the registered word a conversion target, and converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information according to a predetermined conversion rule for other words. Here, the game situation refers to all information that represents the situation of the video game related to the application of the registered word. Examples of information that represent the game situation include the user's playing time, the progress of the story, the level of the user character, information on which character the user uses, and the situation of the relationship with each NPC.

[0051] As a premise for the processing in the conversion unit 14Z, a registered word to be applied is registered for each NPC and / or for each game situation. That is, the registered word to be applied may be set to be different depending on the NPC, and the registered word to be applied may be set to be different depending on the game situation. This makes it possible to express the growth of an NPC according to the progress of the game by setting the registered words so that the answer sentence for a young NPC is all converted to hiragana, and as the game situation progresses and the NPC grows, the answer sentence gradually uses kanji. Also, in a situation where the NPC is a foreign NPC and language comprehension is uncertain, an answer sentence is generated using katakana or a unique character string (which may be an unreadable character string), but in order to express that the NPC's language comprehension has progressed as the game situation progresses, it is possible to make the answer sentence gradually use more hiragana and kanji.

[0052] As an example of changing the registered words to be applied depending on the game situation, it is possible to change the registered words depending on the progress of the game. For example, in the tutorial immediately after the start of a video game, the number of registered words to be applied can be set to be fewer so that there are more hiragana characters, and in the main part of the video game, the number of registered words to be applied can be increased and more kanji characters can be displayed. In this way, it is possible to easily change the atmosphere of the game between the tutorial and the main part.

[0053] As an example of changing the registered words applied depending on the game situation, in a situation where a user can use multiple characters, when the character placed at the beginning (i.e. the character who talks with the NPC) is changed, the content of what the NPC says will be the same, but the expression of the converted sentence will change by changing the registered words applied. For example, in a game where a party of four characters goes on an adventure, the atmosphere in which the NPC speaks can be changed depending on which character is placed at the beginning, making it possible to improve the entertainment value of the game.

[0054] Another example of applying different registered words depending on the game situation is to apply different registered words depending on the physical condition or status of the character. For example, when the character's status is normal, many registered words are applied to create sentences using many kanji characters, but when the character's status is confused, the registered words are not applied and the text is converted to hiragana, and when the character's status is minimum (dwarf state), the registered words are not applied and the text is converted to katakana. This makes it possible to express the character's status in sentences.

[0055] As an example of changing the registered words to be applied depending on the game situation, it is possible to change the registered words to be applied randomly (to make them appear to change randomly). By changing the registered words to be applied each time, it is possible to produce an effect such as an NPC who is psychologically unstable and has an unstable way of speaking, or an NPC of a sub-human race who is highly intelligent and has correct grammar and vocabulary but cannot pronounce words correctly. Any method can be used as long as it makes it appear to change randomly. The application or non-application of each registered word may be determined randomly or based on a variable, or multiple groups related to combinations of registered words to be applied may be set, and which group to apply may be determined randomly or based on a variable.

[0056] In addition, the conversion unit 14Z may be able to change the registered words to be applied based on a user operation. For example, it is possible to select kanji expected to be read for each grade level and set them as registered words of non-replacement words, such as a dictionary for first graders (dictionary for 7-year-olds), , and a dictionary for sixth graders (dictionary for 12-year-olds), and the user may select a dictionary to be applied in advance.

[0057] In addition, the conversion unit 14Z converts a portion that is not a registered word into either hiragana or katakana. The choice of whether to convert into hiragana or katakana may be a user-selectable item, or may be set for each NPC and / or each game situation in the video game. In this case, when the setting is to convert into hiragana, the reserved word set to be replaced with katakana as a registered word becomes effective, and when the setting is to convert into katakana, the reserved word set to be replaced with hiragana as a registered word becomes effective. For example, when the conversion unit 14Z is set to convert into hiragana, the process of changing "ore" to "ore" is realized by the conversion process in the conversion unit 14Z, but in order to convert "ore" to "ore" in this setting, it is necessary to register it as a reserved word. In other words, characters different from the base conversion characters are expressed by applying reserved words.

[0058] The processing may be shifted to the reading information acquisition unit 13Z and the conversion unit 14Z after the generation of the answer sentence in the answer sentence generation unit 12Z is ​​completely finished, but the processing may be executed in the reading information acquisition unit 13Z and the conversion unit 14Z sequentially from a part of the generated sentence during the generation of the answer sentence in the answer sentence generation unit 12Z. In other words, the converted answer sentences may be dynamically generated sequentially during the generation of the answer sentence.

[0059] Next, the operation of the information processing system 100 of this embodiment will be described.

[0060] 9 is a flowchart showing an example of information processing corresponding to at least one of the embodiments of the present invention. Hereinafter, an example will be described in which the server 10Z and the user terminal 20 execute information processing.

[0061] The information processing is started by the user terminal 20 transmitting an input sentence to the server 10Z (step S31). The server 10Z acquires the input sentence from the user terminal 20 (step S32). Next, the server 10Z generates an answer sentence using a language model (step S33). Next, the server 10Z acquires reading information for the answer sentence (step S34). Next, the server 10Z identifies an NPC with which the user will converse, and / or identifies a registered word to be applied by referring to the game situation when the input sentence was acquired from the user terminal, and executes a process set for each registered word for the registered word included in the answer sentence, and converts the other kanji characters to hiragana (step S35). Next, the server 10Z outputs the converted answer sentence (step S36). Then, the user terminal 20 receives the converted answer sentence (step S37), and ends the information processing.

[0062] 10 is a flowchart showing an example of the operation on the server side in information processing corresponding to at least one of the embodiments of the present invention. Here, the operation of the server 10Z in the information processing system 100 will be described again.

[0063] The information processing is started by the server 10Z acquiring an input sentence from the user terminal 20 (step S301). Next, the server 10Z generates an answer sentence using a language model (step S302). Next, the server 10Z acquires reading information for the answer sentence (step S303). Next, the server 10Z identifies an NPC with which the user will converse, and / or identifies registered words to be applied by referring to the game situation when the input sentence was acquired from the user terminal, and executes processing set for each registered word for registered words included in the answer sentence, and converts other kanji characters to hiragana (step S304). Then, the server 10Z outputs the converted answer sentence (step S305), and ends the information processing.

[0064] 11 is a flowchart showing an example of the operation of the terminal side in information processing corresponding to at least one of the embodiments of the present invention. Hereinafter, an example will be described in which the user terminal 20 executes information processing by itself. Note that the configuration of the user terminal 20 has the same functions as the configuration of the server 10Z except for receiving various information from the server 10Z, so a description thereof will be omitted to avoid duplication.

[0065] The information processing is started by acquiring an input sentence in the user terminal 20 (step S401). Next, the user terminal 20 generates an answer sentence using a language model (step S402). Next, the user terminal 20 acquires reading information for the answer sentence (step S403). Next, the user terminal 20 identifies an NPC with which the user will converse, and / or identifies registered words to be applied by referring to the game situation when the input sentence was acquired from the user terminal, executes processing set for each registered word for registered words included in the answer sentence, and converts other kanji characters to hiragana (step S404). Then, the user terminal 20 outputs the converted answer sentence (step S405), and ends the information processing.

[0066] FIG. 12 is an explanatory diagram for explaining an example of conversion processing of an answer sentence in information processing corresponding to at least one of the embodiments of the present invention. In FIG. 12, the sentence "Boss! I have received a message from Kyoko, Tsukimisato's daughter" is given as an example of an answer sentence before conversion processing. Two examples of conversion processing for this example sentence are given. Pattern 1 is an example of conversion to hiragana. When the example sentence is converted to hiragana, it becomes "Boss! I have received a message from Kyoko, Tsukimisato's daughter." In this case, "Tsukimisato" may be registered as a reserved word to be replaced with "Yamanashi" and processed. Also, pattern 2 is an example of registering proper nouns and katakana as non-replaceable words and retaining them as they are. When converted with pattern 2, it becomes "Boss! I have received a message from Kyoko, Tsukimisato's daughter." In this way, various conversion processes can be applied to the same answer sentence depending on the pattern of use of the registered words.

[0067] As described above, as one aspect of the third embodiment, the server 10Z is equipped with an input sentence acquisition unit 11Z, an answer sentence generation unit 12Z, a reading information acquisition unit 13Z, and a conversion unit 14Z, and acquires an input sentence from a user terminal, inputs the input sentence into a predetermined language model to output an answer sentence, analyzes the answer sentence using a predetermined analysis means to acquire reading information, and converts at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with predetermined conversion rules. In the conversion, if a pre-registered word is included in the answer sentence, the registered word is not subject to conversion and a pre-set process is executed for the registered word. Therefore, by executing the process set for each registered word and converting words other than the registered word into hiragana or katakana, the answer sentence generated by the language model can be converted into one that appropriately includes kanji and pre-registered character strings.

[0068] In other words, when the answer sentence generated by the language model is converted into hiragana or katakana after the fact, registered words such as non-substitution words, which are words that use kanji characters as they are, and reserved words that are converted into predetermined character strings, are extracted and processing is performed for each registered word, making it possible to convert the answer sentence into the desired state.

[0069] Furthermore, as one aspect of the third embodiment, the conversion unit 14Z identifies the NPC with which the user is speaking and / or identifies the registered word to be applied by referring to the game situation at the time the input sentence is obtained from the user terminal, and if the registered word to be applied is included in the answer sentence, the registered word is not subject to conversion and a preset process is executed for the registered word, and for other words, at least a portion of the kanji in the answer sentence is converted to hiragana or katakana based on the reading information in accordance with predetermined conversion rules. Therefore, it is possible to change the registered word to be applied and the conversion content of the answer sentence depending on who the NPC with which the user is speaking is and what the game situation is.

[0070] In other words, by setting different registered words for each NPC and each game situation, it is possible to reflect the personality of the NPC or the game situation, thereby increasing the sense of immersion in the video game without the player feeling that the answers are coming from a language model.

[0071] In the first to third embodiments, the content of the answer sentence generated by the language model is not changed. However, the present invention is not limited to this. A part of the expression of the answer sentence generated by the language model may be changed. For example, a registered word for changing the expression of a specific word, the expression at the end of a sentence, etc. to a specific expression may be registered in advance to change a part of the expression of the answer sentence generated by the language model to express the dialect or catchphrase of the NPC. In addition, a part of the expression of the answer sentence generated by the language model may be changed using a registered word to express the physical condition or mood of the NPC. As an example, a registered word may be set as a reserved word for replacing the period "." with "..." to correspond to a case where the NPC is tired or depressed as a game situation, and a registered word may be set as a reserved word for replacing the period "." with "!" to correspond to a case where the NPC is in perfect health or in high spirits as a game situation.

[0072] In the first to third embodiments, the conversion unit 14 converts kanji into hiragana or katakana as a rule, and registered words are used to express exceptional processing, but this is not limited to this. That is, the processing may be reversed so that the conversion unit 14 does not convert kanji into hiragana or katakana as a rule, but only words registered as replacement words are converted into hiragana or katakana. Even in this case, it is possible to separately register registered words as reserved words and apply them.

[0073] In the first to third embodiments, the conversion unit 14 converts kanji into hiragana or katakana in principle. While converting kanji into hiragana or katakana is useful for people who cannot read kanji, a string of hiragana or katakana can be difficult to read. Therefore, when converting into hiragana or katakana, spaces may be inserted between words. As a specific method, the conversion unit 14 divides the text into tokens by morphological analysis, and the conversion unit 14 processes the text so that spaces are always inserted between the tokens. In addition, as a method for making a specific word easier to read, when reserving the replacement of kanji with hiragana as a reserved word, the entire string of hiragana characters corresponding to the kanji characters with spaces inserted before and after them may be registered as a reserved word, so that spaces are inserted when the reserved word is replaced.

[0074] In the first to third embodiments, the kanji included in the answer sentence is converted to hiragana or katakana as a basic process, but the present invention is not limited to this. For example, the process up to the extraction of reading information about the answer sentence is performed in the same manner as in the first to third embodiments, and the answer sentence generated by the language model is presented to the user as is. If the user wishes, for example, if some switching operation is performed, the kanji part in the answer sentence may be converted to hiragana or katakana, or the kanji part in the answer sentence may be displayed with furigana. In this case, the display may be switched immediately by displaying the kanji while internally caching the hiragana.

[0075] In the first to third embodiments, the language model generates an answer sentence for the input sentence, and then processes and converts the registered words. However, a similar process may be applied to a sentence prepared in advance. For example, various sentences may be registered as candidates for the lines of NPCs in a video game, and when the sentences are displayed as the lines of any NPC, the sentences may be converted using the registered words for each NPC or the registered words according to the game situation before being displayed. In other words, it is possible to reflect the personality of the NPC in the video game or the game situation, not limited to the sentences generated by the language model, and to enhance the immersion in the video game.

[0076] As described above, each embodiment of the present application solves one or more deficiencies. Note that the effects of each embodiment are non-limiting effects or examples of effects.

[0077] In each of the above-described embodiments, the multiple user terminals 20, 201 to 20N and the server 10 execute the above-described various processes according to various control programs (for example, information processing programs) stored in their own storage devices.

[0078] Furthermore, the configuration of the information processing system 100 is not limited to the configurations described as examples of each of the above-mentioned embodiments, and may be configured such that the server 10 executes some or all of the processes described as processes executed by the user terminal 20, or such that any one of the multiple user terminals 20, 201-20N (e.g., the user terminal 20) executes some or all of the processes described as processes executed by the server 10. Also, any one of the multiple user terminals 20, 201-20N may have some or all of the storage unit included in the server 10. In other words, the system 100 may be configured such that some or all of the functions included in either the user terminal 20 or the server 10 are included in the other one.

[0079] Furthermore, the program may be configured to realize some or all of the functions described as examples of each of the above-mentioned embodiments in a standalone device that does not include a communication network.

[0080] [Note] The above-mentioned embodiment has been described so as to enable at least one person having ordinary skill in the art to which the invention pertains to practice the invention. [1] An information processing program for causing a server to realize a process of generating an answer using a dialogue AI for an input sentence from a user terminal, The server, an input sentence acquisition function for acquiring the input sentence from the user terminal; an answer sentence generation function that inputs the input sentence into a predetermined language model and outputs an answer sentence; a reading information acquisition function for acquiring reading information by analyzing the answer sentence by a predetermined analysis means; A conversion function for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; The conversion function performs a preset process on a registered word when the reply sentence contains a registered word that has been registered in advance, without converting the registered word. Information processing program. [2] An information processing program for enabling conversations between users and non-player characters (hereafter referred to as NPCs) within a video game, The registered words are registered for each of the NPCs and / or for each game situation; The conversion function identifies the NPC with which the user is to converse, and / or identifies the registered word to be applied by referring to the game situation when the input sentence is acquired from the user terminal. [1] An information processing program as described above. [3] The registered words are reserved words that are to be replaced with preregistered character strings. An information processing program according to [1] or [2]. [4] The registered words are non-conversion words that use kanji as is without converting them to hiragana or katakana. An information processing program according to any one of [1] to [3]. [5] An information processing program that causes a user terminal capable of communicating with the server to realize at least one of the functions that the information processing program described in any one of [1] to [4] causes the server to realize. [6] An information processing system comprising a communication network, a server, and a user terminal, for performing a process of generating an answer using a dialogue AI in response to an input sentence from the user terminal, an input sentence acquisition means for acquiring the input sentence from the user terminal; an answer sentence generating means for inputting the input sentence into a predetermined language model and outputting an answer sentence; a reading information acquiring means for acquiring reading information by analyzing the answer sentence by a predetermined analyzing means; a conversion means for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; When a pre-registered word is included in the answer sentence, the conversion means performs a pre-set process on the registered word without converting the registered word. Information processing system. [7] A server for executing a process of generating an answer using a dialogue AI for an input sentence from a user terminal, an input sentence acquisition unit that acquires the input sentence from the user terminal; an answer sentence generation unit that inputs the input sentence into a predetermined language model and outputs an answer sentence; a reading information acquisition unit that analyzes the answer sentence by a predetermined analysis means and acquires reading information; a conversion unit that converts at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; When a pre-registered word is included in the answer sentence, the conversion unit performs a pre-set process on the registered word without converting the registered word. server. [8] An information processing program for causing a user terminal to realize a process of generating an answer to an input sentence using a dialogue AI, The user terminal, an input sentence acquisition function for acquiring the input sentence; an answer sentence generation function that inputs the input sentence into a predetermined language model and outputs an answer sentence; a reading information acquisition function for acquiring reading information by analyzing the answer sentence by a predetermined analysis means; A conversion function for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; The conversion function performs a preset process on a registered word when the reply sentence contains a registered word that has been registered in advance, without converting the registered word. Information processing program. [9] An information processing method for performing a process of generating an answer to an input sentence using a dialogue AI, comprising: an input sentence acquisition step for acquiring the input sentence; an answer sentence generation step of inputting the input sentence into a predetermined language model and outputting an answer sentence; a reading information acquisition step of analyzing the answer sentence by a predetermined analysis means to acquire reading information; A conversion procedure for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; The conversion step is to execute a preset process for a registered word when the reply sentence contains a registered word that has been registered in advance, without converting the registered word. Information processing methods. [Industrial Applicability]

[0081] According to one embodiment of the present invention, conversion processing using registered words makes it possible to express the individuality of a pseudo-personality on the AI ​​side in a dialogue-based AI service. [Explanation of symbols]

[0082] 10 Server 11 Input text acquisition section 12 Answer text generation section 13 Reading information acquisition unit 14 Conversion section 20,201~20N User terminals 30 Communication Network 100 Information Processing Systems

Claims

1. An information processing program for causing a server to realize a process of generating an answer using an interactive AI for an input sentence from a user terminal, The server, an input sentence acquisition function for acquiring the input sentence from the user terminal; an answer sentence generation function that inputs the input sentence into a predetermined language model and outputs an answer sentence; a reading information acquisition function for acquiring reading information by analyzing the answer sentence by a predetermined analysis means; A conversion function for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; The conversion function performs a preset process on a registered word when the reply sentence contains a registered word that has been registered in advance, without converting the registered word. Information processing program.

2. An information processing program for enabling a conversation between a user and a non-player character (hereinafter referred to as NPC) in a video game, The registered words are registered for each NPC and / or each game situation; The conversion function identifies the NPC with which the user will converse, and / or identifies the registered word to be applied by referring to the game situation when the input sentence is acquired from the user terminal. The information processing program according to claim 1.

3. A server for executing a process of generating an answer using a dialogue type AI for an input sentence from a user terminal, an input sentence acquisition unit that acquires the input sentence from the user terminal; an answer sentence generation unit that inputs the input sentence into a predetermined language model and outputs an answer sentence; a reading information acquisition unit that analyzes the answer sentence by a predetermined analysis means and acquires reading information; a conversion unit that converts at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; When a pre-registered word is included in the answer sentence, the conversion unit performs a pre-set process on the registered word without converting the registered word. server.

4. An information processing system comprising a communication network, a server, and a user terminal, for performing a process of generating an answer using a dialogue type AI in response to an input sentence from the user terminal, an input sentence acquisition means for acquiring the input sentence from the user terminal; an answer sentence generating means for inputting the input sentence into a predetermined language model and outputting an answer sentence; a reading information acquiring means for acquiring reading information by analyzing the answer sentence by a predetermined analyzing means; a conversion means for converting at least a part of the kanji in the answer sentence into hiragana or katakana based on the reading information in accordance with a predetermined conversion rule; When a pre-registered word is included in the answer sentence, the conversion means performs a pre-set process on the registered word without converting the registered word. Information processing system.

Images