JP6644141B2 - Response device, control method of response device, and control program - Google Patents

Description

  The present invention relates to a response device that recognizes input contents and responds based on the recognized contents.

  2. Description of the Related Art A response device that receives an utterance of a user, recognizes input content from the utterance, and responds according to the recognized result is known as a related art. For example, Patent Document 1 below discloses a conversation scenario editing device that generates and edits a conversation scenario used in an automatic conversation device that is a device capable of establishing a conversation with a user.

Japanese Unexamined Patent Publication "JP-A-2010-73192" (published on April 2, 2010)

  However, the above-described conventional technology is configured to respond to a recognized result based on a conversation scenario prepared in advance, but the number of scenarios is limited. Therefore, if no conversation scenario that matches the recognized result is found, only the contents specified in advance in <Others> can be answered. Furthermore, if the number of scenarios is small, the response contents will be similar, resulting in a problem that the conversation will be boring.

  The present invention has been made in view of the above problems, and its purpose is to provide a response message based on a phrase included in an input content even when a conversation scenario matching the input content cannot be found. To implement a response device that generates

  In order to solve the above-described problem, a response device according to one embodiment of the present invention is a response device that outputs a response message for an input content input by a user, wherein a first phrase and a first related phrase are related to each other. And a response message generation unit that generates a response message to the input content, the storage unit storing that the second phrase and the second related phrase are related to each other, The first word or the second word is included in the input content, and the storage unit stores that the second word and the first related word are related to each other, or that the first word and the first word are related to each other. When the fact that the second word and phrase belong to the same concept is stored, the response message generation unit generates the response message having a content in which the first word and the second related word are associated with each other.

  Further, in order to solve the above problem, a control method of a response device according to one embodiment of the present invention is a control method of a response device that outputs a response message for an input content input by a user, wherein the response device is And a storage unit for storing that the first phrase and the first related phrase are related to each other, and that the second word and the second related phrase are related to each other. A search step of searching the storage section for the included first or second phrase, and a response message generation step of generating a response message to the input content, wherein the first or second phrase is included. A word is included in the input content, and the storage unit stores information that the second word and the first related word are related to each other, or that the first word and the second word are the same. If that stores that belong to the concept, in the response message generation step, generating a response message content associated with each other and the first word and the second related phrases.

  Advantageous Effects of Invention According to one aspect of the present invention, it is possible to realize a response device that generates a response message based on a phrase included in an input content even when a conversation scenario that matches the input content cannot be found. .

FIG. 2 is a block diagram illustrating an example of a configuration of a main part of the robot 1 according to the first embodiment of the present invention. 4 is an example of phrase information provided in a storage unit of the robot 1 according to the first embodiment of the present invention. FIG. 3 is a diagram illustrating an example of an input / output history provided in a storage unit of the robot 1 according to the first embodiment of the present invention. 4 shows an example of a response message template provided in the storage unit of the robot 1 according to the first embodiment of the present invention, and an example sentence using the template. FIG. 5 is a diagram illustrating an image in which the robot 1 according to the first embodiment of the present invention generates a response message by inference. FIG. 6 is a diagram illustrating another image in which the robot 1 according to the first embodiment of the present invention generates a response message by inference. 4 is a flowchart illustrating an example of a process executed by the robot according to the first embodiment of the present invention. FIG. 13 is a diagram illustrating another image in which the robot 1 according to the second embodiment of the present invention generates a response message by inference. FIG. 13 is a diagram illustrating another image in which the robot 1 according to the second embodiment of the present invention generates a response message by inference.

[Embodiment 1]
Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS.

<Robot configuration>
An outline of the robot 1 according to the present embodiment will be described based on FIG. Note that the robot 1 according to the present embodiment is a response device that outputs a response message with respect to the input content input by the user by speaking. FIG. 1 is a block diagram illustrating an example of a main configuration of the robot 1 according to the present embodiment. As shown in FIG. 1, the robot 1 includes a voice input unit 11, a voice output unit 12, a storage unit 13, and a control unit 20. The voice input unit 11 detects a user's utterance. The sound input unit 11 may be any sound collecting device such as a microphone. The voice input unit 11 sends the detected utterance of the user as voice data to a voice recognition unit 21 described later. Note that the voice input unit 11 specifies one utterance (a utterance that forms a single sentence or sentence) from the time of the user's utterance (time during which no voice is being emitted) and the like, and outputs the voice for each utterance. It is desirable to transmit the data to the voice recognition unit 21. The voice output unit 12 functions as an output unit that outputs voice data received from a voice synthesis unit 27 described later as voice to the outside. The audio output unit 12 is specifically realized by a speaker or the like provided in the robot 1. In the example of FIG. 1, the audio output unit 12 is built in the robot 1, but the audio output unit 12 may be an external device attached to the robot 1. The storage unit 13 stores various data handled by the robot 1. The storage unit 13 according to the present embodiment stores at least a phrase information, an input / output history, and a template of a response message described later.

<Example of data stored in storage unit>
An example of data stored in the storage unit 13 in the present embodiment will be described with reference to FIGS. FIG. 2 is an example of phrase information stored in the storage unit 13 by the robot 1 according to the present embodiment, and FIG. 3 is an input / output history stored in the storage unit 13 by the robot 1 according to the present embodiment. This is an example. FIG. 4 shows an example of a response message template stored in the storage unit 13 by the robot 1 according to the first embodiment of the present invention, and an example sentence using the template.

  In the example of FIG. 2, the phrase information stored in the storage unit 13 manages the attribute to which the phrase belongs and the relevance to other phrases, and includes items of the phrase, concept, and related phrase. , Parts of speech, classification, and usage status. In the illustrated example, the term indicates not only a noun such as “apple” and “Osaka”, but also a word having a single meaning, including adjectives such as “sweet” and “round” and a verb “buy”. Words such as particles that do not make sense by themselves are not included. The part of speech indicates the type of part of speech to which each word belongs, such as “noun”, “adjective”, “verb”, and the like. The classification indicates a broader category to which each word belongs, such as "fruit" or "domestic region". The usage status indicates a status in which each word is to be used, such as “What (what)” and “Where (where)”. That is, if the combination of concepts (part of speech, classification, usage status) is the same for a plurality of words, the response message generation unit 26 can determine that those words belong to the same concept. The related phrase indicates, for each phrase, another phrase that is associated with the phrase and that can be combined with the phrase and / or a template of the response message to form a response message. That is, the storage unit 13 stores that each word is related to each other via the related word. In the illustrated example, since the related phrase of “apple” is “sweet”, these can be combined to form the response message “apple is sweet”.

  In the example of FIG. 3, the input / output history provided in the storage unit 13 manages history information related to input / output of the robot 1 and includes items of a number, a date, an input / output, a content, and a used phrase. The number manages the input content of the user and the output content of the robot 1 in chronological order, and the date and time records the date and time when the user or the robot 1 performs input / output. The input / output indicates whether the history information is from the user or the robot 1. In the illustrated example, “input” indicates a user input, and “output” indicates an output of the robot. The contents indicate the contents of input / output by the user or the robot 1. The use phrase indicates a phrase included in the input / output content of the user or the robot 1. In the illustrated example, up to three terms 1 to 3 are recorded, but the present invention is not limited to this.

  FIG. 4 shows a template of the response message provided in the storage unit 13, its outline, and an example sentence using the template. In the illustrated example, the storage unit 13 only needs to store at least items of the template, and does not need to store items of the outline and the example sentences. The template is a template of a sentence to be selected by the robot 1 and combined with a word to compose a response message. In the illustrated example, portions where the word is incorporated are described as [A] and [B]. For example, in the first template, "[A] looks like [B], why?", A response message is formed by replacing each of [A] and [B] with a phrase. The outline indicates the content of the response message corresponding to the template. In the example shown in the figure, the outline of the template "[A] is not [B] why is it?" Is "Joke". The example sentence shows an example of a response message in a case where a phrase acquired from the input content of the user is combined with each template. In the illustrated example, the phrase replaced by [A] is “Kinkakuji”, and the phrase replaced by [B] is “sweet”. At this time, as can be seen from the illustrated example, the template includes information indicating how related the phrase replaced by [A] and the phrase replaced by [B] are. For example, the relevance is represented by word order, particles, or the like. In addition, if there is no fixed sentence applicable to the plurality of templates, the robot 1 may generate, as a response message, the phrase "I wondered how I heard" written at the end of the figure.

  The control unit 20 controls the components of the robot 1 collectively, and includes a voice recognition unit 21, a phrase / concept acquisition unit 22, a scenario unit 23, a concept determination unit 24, a related phrase acquisition unit 25, and a response message generation unit. 26, and a speech synthesizer 27.

  The voice recognition unit 21 performs voice recognition on voice data of one utterance received from the voice input unit 11. In the present invention, “speech recognition” refers to a process of obtaining text data indicating utterance contents (input contents) from utterance voice data. The method of voice recognition of the voice recognition unit 21 is not particularly limited, and voice recognition may be performed using any conventional method.

  When receiving the text data obtained by the speech recognition unit 21 by the speech recognition, the phrase / concept acquisition unit 22 divides the text data into phrases, and acquires the concept of each divided phrase from the storage unit 13. Further, the phrase / concept acquisition unit 22 sets a first phrase (or a second phrase) and a first related phrase (or a second related phrase), which will be described later, from the separated phrases. Furthermore, when each of the separated phrases does not exist in the phrase information of FIG. 2, the phrase / concept acquisition unit 22 may newly register the phrase in the phrase information. If one of the separated words is not registered as a related word of another separated word, the word may be registered as a related word of another word.

  The scenario unit 23 obtains, from the scenario information stored in the storage unit 13, scenario information corresponding to the word and the concept of the word acquired by the word / concept acquisition unit 22.

  The concept determination unit 24 determines whether the concept of the phrase acquired by the phrase / concept acquisition unit 22 is the same as the concept of the phrase included in the output content immediately before the robot 1 acquired from the storage unit 13. judge. That is, the concept determining unit 24 determines whether the content input by the user is a content corresponding to the response message output immediately before the input.

  The related phrase acquiring unit 25 associates the related phrase with the phrase based on the phrase and the concept of the phrase acquired by the phrase / concept acquiring unit 22 and separates the related phrase that is different from the phrase into the phrase information in FIG. To get from. Further, the related phrase acquiring unit 25 sets the acquired related phrase as a phrase to be used for inference described later. Here, the condition for the related phrase acquisition unit 25 to acquire the related phrase may be any condition. For example, for a specific phrase, all the phrases registered as related phrases in the phrase information of FIG. 2 may be acquired, only the phrase not included in the input content may be acquired, or the concept of the phrase may be acquired. A phrase whose (part of speech) is "noun" may be preferentially acquired.

  The response message generator 26 generates text data serving as a response message output by the robot 1 in accordance with the input content of the user, based on the instruction content from each unit. Specifically, at least one of the phrase acquired by the phrase / concept acquisition unit 22 and the related phrase acquired by the related phrase acquisition unit 25, the template stored in the storage unit 13, the history information, and the scenario unit 23 acquire it. Text data is generated by combining any of the scenario information as needed. Here, when there are a plurality of related phrases acquired by the related phrase acquiring unit 25, the response message generating unit 26 may be configured to extract a specific related phrase according to a predetermined condition. For example, the response message generation unit 26 is configured to preferentially extract related words belonging to the same concept as the words and phrases acquired by the word / concept acquisition unit 22 and generate a response message having contents in which these words are associated with each other. You may. Note that the response message generator 26 may be of any configuration as long as it generates text data serving as a response message. For example, like the phrase / concept acquisition unit 22, the response message generation unit 26 determines, for a plurality of phrases included in the input content of the user, based on context information including the concept to which each phrase belongs and the phrases before and after each phrase. May be further provided to select the input word. Here, the context information is information indicating whether a phrase is classified into a subject, a predicate, or any other of the input content. For example, the response message generator 26 may be configured to preferentially select, as an input phrase, a phrase whose concept (classification) belongs to a proper noun and a phrase that is a predicate in the input content.

  The voice synthesizer 27 converts the text data of the response message received from the response message generator 26 into voice data. The voice synthesis unit 27 outputs the converted voice data to the voice output unit 12.

<Robot operation>
An outline of the operation of the robot 1 according to the present embodiment will be described based on FIG. FIG. 5 shows an image in which the robot 1 according to the present embodiment generates a response message by inference. In the following description, it is assumed that the user has made an input to the robot 1 by saying “Apple is sweet”.

  First, when the user speaks to the robot 1 "Apple is sweet", the robot 1 converts the speech into text data by voice recognition, and divides the text data into phrases. That is, if "apple is sweet", it is divided into three words "apple", "ha", and "sweet". At this time, the robot 1 may register or update each of the separated words in the word information in FIG.

  Next, the robot 1 selects a word to be a starting point of the inference from the separated words, and sets the selected word as a first word (or a second word). Here, the first word (or the second word) is an input word that is a word characterizing the input content of the user. Note that the first word (or the second word) may be set to any word as long as it has a single meaning, similarly to the word information shown in FIG. 2, and may be set according to a predetermined priority. It is preferable to set. For example, when the predicate phrase includes a noun, this is set as the first phrase (or second phrase), and when not, the noun that is the subject is set as the first phrase (or second phrase). It may be. In the case where noun is not included in either the subject or the predicate, an appropriate phrase may be set as the first phrase (or the second phrase). According to the above example, among the “apple”, “ha”, and “sweet”, the predicate “sweet” is an adjective, and the subject “apple” is a noun. Is set as the first phrase (or the second phrase). Further, among words that are not set as the first word (or the second word), a word that is recognized as being related to the first word (or the second word) is first related word (or the second word). Related terms). According to the above example, “sweet” is set as the first related phrase (or the second related phrase).

  After setting the first word (or the second word) and the first related word (or the second related word), the robot 1 sets the first related word (or the second related word) according to the word information in FIG. A second phrase (or first phrase), which is a related phrase, is acquired and set. According to the above example, the related terms of the first related word (or the second related word) “sweet” are “apple” and “melon”, and “apple” is the first word (or the second word). Term). Therefore, here, "melon" is set as the second word (or the first word).

  When the second phrase (or the first phrase) is set, the robot 1 according to the phrase information of FIG. 2, the second related phrase (or the first related phrase), which is a related phrase of the second phrase (or the first phrase). ) To get and set. It is preferable that the concept (usage status) of the second related phrase (or the first related phrase) is “How”. This is because when the concept of the second related phrase (or the first related phrase) is not “How”, a natural sentence is not often obtained when combined with the second related phrase (or the first related phrase). That's why. According to the above example, the related phrases of the second word (or the first word) “melon” are “sweet”, “round”, and “buy”. Here, since “sweet” has already been set as the first related term (or the second related term), the remaining “round” or “buy” can be set as the second related term (or the first related term). is there. At this time, a comparison between “round” and “buy” using the phrase information in FIG. 2 shows that the concept (usage status) of “round” is “How (how)” and the concept of “buy” ( Usage status) is "-". In this case, the robot 1 sets “round” as the second related phrase (or the first related phrase).

  Then, the robot 1 (the response message generation unit 26), as in S23 in FIG. 7 described later, acquires the first word (or the second word) and the second related word (or the first related word) acquired by the processing so far. From the phrase).

  An example of a method for generating a response message will be described below. In the following example, the phrase information of FIG. 2 also stores particles that do not make sense only with the phrase such as particles. Specifically, the phrase information in FIG. 2 stores a phrase “ha”. At this time, the combination of the concept (part of speech, classification, usage) of the word "ha" is (particle,-,-), and the related words are "apple", "melon", "sweet", and "round". It is. In the above example, the first phrase (or the second phrase) is “apple”, and the second related phrase (or the first related phrase) is “round”. Further, the second phrase (or the first phrase) is “melon”. Therefore, the robot 1 uses "apple" and "round", and further generates "response is round" as a response message in combination with a particle "ha" having both "apple" and "melon" as related words. In this manner, the robot 1 can generate a response message by inference based on the words included in the input content.

  The configuration for generating the response message is not limited to the above. Another example of a configuration for generating a response message will be described below. First, when the robot 1 acquires the second word (or first word) and the second related word (or first related word), the history information including these words is obtained from the input / output history shown in FIG. To get. The history information includes information indicating how the second word (or the first word) is related to the second related word (or the first related word). For example, the relevance is represented by word order, particles, or the like. According to the above example, the history information whose “content” is “melon is round” is acquired. Then, the robot 1 (the response message generation unit 26) replaces the second phrase (or the first phrase) included in the “contents” of the history information with the first phrase (or the second phrase), and thereby the response message is generated. Generate That is, in the above example, “apple is round” is generated as a response message. By using the “contents” of the input / output history, the robot 1 can appropriately determine the word order of each word in the response message and appropriately combine words such as particles that do not make sense with the word alone. Note that the storage unit 13 may store information on relevance between a plurality of phrases (word order and / or particles between the plurality of phrases) separately from the history information.

  When the second phrase (or the first phrase) is not found, the robot 1 (the response message generation unit 26), as in S24 of FIG. 7 described later, sets the first related phrase (or the second related phrase) and A configuration in which a response message is generated from the template in FIG. 4 may be used. In the above example, for example, if the template is "What is there more when you say [A]?", "Sweet and sweet" combined with the first related word (or second related word) "Sweet" Speaking what else? "Is generated as a response message. At this time, it is expected that the user's input content for the response message includes a second phrase (or first phrase) associated with the first related phrase (or second related phrase). Therefore, the robot 1 is configured to register, in the phrase information of FIG. 2, the phrase acquired from the input content of the user with respect to the response message, as the related phrase of the first related phrase (or the second related phrase) used in the response message. There may be. According to the above configuration, it is possible to acquire a related phrase from the user's input content for the response message, so that the convenience of the robot 1 can be improved.

  Further, when the input content of the user ignores the flow of the dialog, the robot 1 generates another response message from the first phrase and the template of FIG. 4 as in S26 of FIG. 7 described later. It may be. At this time, the robot 1 may have any configuration as long as the content of the user's input can be compared with the flow of the dialog. For example, the robot 1 (the concept determination unit 24) acquires the content output by the robot 1 immediately before the user's input from the input / output history of FIG. 3, and sets the words included in the history as the third words. At this time, the robot 1 determines whether or not the concept to which the third word belongs and the concept to which the first word (or the second word) belongs are the same using the word information shown in FIG. May determine that the input has been performed ignoring the flow of the dialogue. At this time, the robot 1 may generate another response message in which, for example, the first word (or the second word) is combined with a template related to the third word. If the robot 1 utters “Apple is sweet” and then utters “Kinkakuji when speaking of Kyoto”, the concept (classification) of the first word (or second word) of “Kinkakuji” is “ On the other hand, "apple", which is the third word, is "fruit" and has a different concept (classification).

  At this time, the robot 1 converts the first word (or the second word) “Kinkakuji” into the template “[A] related to the“ fruit (food) ”which is the third word“ apple ”concept (classification). "Yes" or "[A] cannot be eaten." That is, according to the example sentence, one of "Kinkakuji is not food" or "Kinkakuji is not eaten" is output as a response message.

  The method of generating another response message is not limited to the above. For example, the robot 1 uses a word different from the first word (or the second word) included in the input content of the user as the fourth word, and converts the first word (or the second word) and the fourth word. , May be combined with the template to generate another response message. If the user utters “Kinkakuji is also sweet” after the robot 1 utters “Apple is sweet,” the concept (classification) of the first word (or second word) “Kinkakuji” and the third word “apple” is But the word "sweet" is common. At this time, the robot 1 uses a word “sweet” different from the first word (or second word) “Kinkakuji” included in the input content of the user as the fourth word. Then, the robot 1 converts the first word (or the second word) and the fourth word into the template “[A] looks like [B], and“ [A] is [B]. Take another one to eat "and" [A] is [B] "to generate another response message. In other words, according to the example sentence, one of "Kinkakuji is sweet isn't it?", "Kinkakuji is sweet, take me to eat next time", or "Kinkakuji is sweet" is output. .

  Also, the inference is not limited to the above-described method, but includes the first phrase (or the second phrase) and the second related phrase (or the first related phrase) necessary for generating a response message in combination with the template. Any method may be used as long as it can be obtained. Another example of the inference will be described with reference to FIG. FIG. 6 shows another image in which the robot 1 according to the present embodiment generates a response message by inference.

  FIG. 6A shows two sentences, “apple is sweet” and “melon is round”. One of the two sentences is the input content, and the other is the past input / output content stored in the storage unit 13. For example, it is assumed that “apple is sweet” is the input content and “melon is round” is the past input / output content. Specifically, the robot 1 receives the input content “apple is sweet” by the user, and sets “apple” as the first phrase. After that, the robot 1 sets the combination of the first word “apple” and the concept (part of speech, classification, usage status) to (noun, fruit, What (what)) and uses the same concept of the second word “melon”. The storage unit retrieves the history information including “melon is round”, and acquires the second phrase “melon”. Since the first phrase “apple” and the second phrase “melon” have the same concept, in this case, the robot 1 maps the two sentences with the same concept as shown in FIG. Generate Then, in an area where (noun, fruit, What (what)) is set, one of “apple” and “melon” is set.

  At this time, from (b) of FIG. 6, four sentences of "apple is sweet", "apple is round", "melon is sweet", and "melon is round" can be obtained. As a result, the robot 1 obtains a sentence “apple is round” and “melon is sweet” containing a combination of two phrases that are not associated with each other in the storage unit 13 as a result of the inference, and generates a response message. Can be.

  The above-described inference may be configured to be used in combination with the generation of a response message based on a conversation scenario, which is a conventional technique. At this time, it is preferable that the storage unit 13 further stores scenario information. In this case, if the scenario information associated with the input content of the user exists in the storage unit 13, the robot 1 performs the first phrase (or the second phrase) included in the input content, as in S25 in FIG. ) And a response message may be generated from the conversation scenario. In addition, when the scenario information does not exist in the storage unit 13, a response message may be generated by inference.

<Process flow>
Next, a flow of processing executed by the robot 1 of the present embodiment will be described with reference to FIG. FIG. 7 is a flowchart illustrating an example of the flow of a process performed by the robot 1 according to the present embodiment. In the following description, it is assumed that the input content of the user includes the first phrase and the first related phrase, and the response message output by the robot 1 immediately before the user inputs the third phrase.

  First, the voice recognition unit 21 performs voice recognition of the input content of the user input received by the voice input unit 11, and generates text data (S11). Next, the phrase / concept acquiring unit 22 receives the text data generated in S11, and extracts the first phrase, the first related phrase, and the first phrase from the text data and the phrase information stored in the storage unit 13. The concept is acquired (S12). Then, the scenario unit 23 searches the storage unit 13 for scenario information corresponding to the first phrase or the concept of the first phrase acquired in S12 (S13), and determines whether the corresponding scenario information is found. (S14). If it is determined that the corresponding scenario information has not been found (NO in S14), the concept determination unit 24 determines whether the robot 1 has responded (uttered) immediately before the user input (S15). On the other hand, when it is determined that the corresponding scenario information has been found (YES in S14), the process proceeds to S25 described later.

  In the processing of S15, when it is determined that the robot 1 has responded (YES in S15), the concept determining unit 24 further determines whether the concept of the first word and the third word is the same (S16). Then, when it is determined that the concepts are the same (YES in S16), the related phrase acquisition unit 25 searches for a second phrase from the related phrase of the first related phrase in the phrase information stored in the storage unit 13 ( S17: Search step). On the other hand, if it is determined that the robot 1 has not responded (NO in S15), the process proceeds directly to S17.

  After S17, the related phrase acquisition unit 25 determines whether a second phrase has been found (whether or not a related phrase of the first related phrase exists in the phrase information) (S18). If it is determined that the second phrase has not been found (NO in S18), the related phrase acquisition unit 25 searches for the second phrase having the same concept as the first phrase from the phrase information stored in the storage unit 13 (S19). ). Then, the related phrase acquisition unit 25 determines whether or not the second phrase has been found in the processing of S19 (S20). If it is determined that the second phrase has been found (YES in S20), or if it has been determined in S18 that the second phrase has been found (YES in S18), the related phrase acquiring unit 25 stores the second phrase in the phrase information in the storage unit 13. A second related phrase is searched from the related phrase of the phrase (S21).

  Thereafter, the related phrase acquisition unit 25 determines whether or not a second related phrase has been found in the process of S21 (whether or not a related phrase of the second phrase exists in the phrase information) (S22). If it is determined that the second related phrase has been found (YES in S22), the response message generator 26 generates a response message from the first phrase and the second related phrase (S23: response message generation step), and the process proceeds to S27. move on. In this way, the response message generation unit 26 generates a response message having contents in which the first phrase and the second related phrase are associated with each other. On the other hand, if the second phrase is not found in the process of S20 (NO in S20), or if the second related phrase is not found in the process of S22 (NO in S22), the response message generation unit 26 sets the first related phrase. A response message is generated from the word and the template (S24). Thereafter, the process proceeds to S27. If it is determined in S14 that the corresponding scenario information has been found (YES in S14), response message generator 26 generates a response message from the first word and the scenario information (S25), and the process proceeds to S27. Then, when it is determined in S16 that the concept of the first word and the third word are not the same (NO in S16), the response message generation unit 26 generates a response message from the first word and the template (S26), and Goes to S27.

  In S27, the voice synthesizing unit 27 converts the response message into voice data and outputs the voice via the voice output unit 12 (S27).

  Through the above processing, the robot 1 according to the present embodiment can generate a response message based on a phrase included in the input content even when a suitable conversation scenario cannot be found.

[Embodiment 2]
Another embodiment of the present invention will be described with reference to FIGS. The robot 1 of the present embodiment is different from the above embodiment in that, when the input content of the user includes three or more words, a response message including three or more words is generated. For convenience of explanation, members having the same functions as the members described in the above embodiment are denoted by the same reference numerals, and description thereof will be omitted. In the robot 1 according to the present embodiment, a first word (or a second word) and a first related word (or a second related word) for three or more words included in the input content of the user in the same manner as in the above embodiment. ) Is set. The phrase / concept acquiring unit 22 is a related phrase of the first related phrase (or the second related phrase). For another phrase different from the first related phrase (or the second related phrase), the first associated phrase (or 2nd accompanying phrase). Here, any criteria may be used for setting the first auxiliary phrase (or the second auxiliary phrase) from the input contents. For example, after the first phrase is set in the same manner as in the above-described embodiment, of the remaining phrases, the phrase whose concept (part of speech) is “part of speech” and whose concept (classification) is other than “others” is first associated. It may be set to a phrase (or a second accompanying phrase). The order in which the first related word (or the second related word) and the first auxiliary word (or the second auxiliary word) are set is not particularly limited, but the phrase whose concept (part of speech) is “noun” is defined as the first word. It is preferable that the setting is made prior to the first accompanying phrase (or the second accompanying phrase). Then, a response message is generated in which the first word (or second word) and the first word (or second word) are associated with the second word (or first word) obtained by inference. .

<Robot operation>
An outline of the operation of the robot 1 according to the present embodiment will be described with reference to FIGS. 8 and 9 are diagrams illustrating an image in which the robot 1 according to the present embodiment generates a response message by inference.

  FIG. 8A shows an image in which inference is applied to two sentences, “Osaka's specialty is okonomiyaki” and “Hiroshima has delicious oysters”. For example, it is assumed that “Hiroshima has delicious oysters” is stored in the storage unit 13 in advance as past inputs and outputs, and “Osaka's specialty is Okonomiyaki” is input by the user. At this time, the robot 1 (phrase / concept acquisition unit 22) displays “okonomiyaki” as the first word (or second word), “specialty” as the first related word (or second related word), and “Osaka” as “ It is set to the first attached phrase (or the second attached phrase). Then, the robot 1 (related phrase acquisition unit 25) includes the phrase “oyster” in which the combination of the first phrase (or the second phrase) “okonomiyaki” and the concept (part of speech, classification, usage status) is the same. The history information “Hiroshima has delicious oysters” including the phrase “Hiroshima” that is the same concept as the accompanying word (or the second accompanying word) “Osaka” is acquired from the input / output history in FIG. Here, "okonomiyaki" and "oysters" are different in that they are classified as "food / okonomiyaki / takoyaki" and "food / seafood", but the portion of "food" is the same. Therefore, it is determined that “okonomiyaki” and “oyster” are words that can be used in the same way (words having different lower-level concepts but matching higher-level concepts). At this time, the robot 1 (response message generating unit 26) generates a map in which two sentences are grouped with the same concept as in FIG. 8B, in the same manner as in FIG. 6B. I do. Then, either “Osaka” or “Hiroshima” is set in the (Noun, Domestic Region, Where (Where)) area, and “Okonomiyaki” and “Okonomiyaki” are set in the (Noun, Food, What (What)) area. Set one of "Oysters". However, since the combination of words to be set often has a meaning, it is necessary to set while maintaining the combination of words used in the same utterance. For example, the first accompanying phrase (or second accompanying phrase) "Osaka" is set in combination with the first (or second) phrase "okonomiyaki", and similarly, "oyster" is set in combination with "Hiroshima". There is a need.

  Therefore, from FIG. 8B, four sentences of “Osaka's specialty is okonomiyaki”, “Osaka is okonomiyaki is delicious”, “Hiroshima's specialty is oysters”, and “Hiroshima is oysters are delicious” are acquired. be able to. As a result, the robot 1 can obtain "Osaka is delicious with okonomiyaki" or "Hiroshima's specialty is oysters" as a result of the inference, and generate a response message.

  FIG. 9A shows an image in which inference is applied to two sentences, “Osaka's specialty is Okonomiyaki” and “Hiroshima souvenirs are maple buns”. For example, it is assumed that “Hiroshima souvenir maple bun” is stored in the storage unit 13 in advance as past input and output, and “Osaka's specialty is okonomiyaki” is input by the user. At this time, the robot 1 (phrase / concept acquisition unit 22) uses “okonomiyaki” as the first word (or second word) and “special product” as the first related word (or second word) in the same manner as in FIG. Related terms) and “Osaka” are set as the first accompanying terms (or the second accompanying terms). Then, the robot 1 (related phrase acquisition unit 25) includes the phrase “momiji bun” in which the combination of the first phrase (or the second phrase) “okonomiyaki” and the concept (part of speech, classification, usage status) is the same. The history information “Hiroshima souvenir maple bun” including the phrase “Hiroshima” that is the same concept as the first accompanying phrase (or the second accompanying phrase) “Osaka” is acquired from the input / output history of FIG. At this time, the robot 1 (response message generation unit 26) generates a map in which two sentences are grouped with the same concept as in FIG. 9B. Here, the combination of the concept (part of speech, classification) between “special product” and “souvenir” matches (noun, other). However, a phrase having a combination of concepts (part of speech, classification) (noun, other) is used without being combined because the feature indicated by the phrase is smaller than other words. In this case, in the same manner as in FIG. 8B, one of “Osaka” and “Hiroshima” is set in the area of (noun, domestic region, Where (where)) from FIG. 9B. I do. Then, in the (Noun, Food, What (what)) area, one of “okonomiyaki” and “momiji bun” is set.

  As a result, the robot 1 can acquire four sentences: “Osaka's specialty is Okonomiyaki”, “Osaka's specialty is Okonomiyaki”, “Hiroshima's specialty is oyster”, and “Hiroshima's specialty is oyster”. As a result, the robot 1 can acquire “Osaka souvenirs are okonomiyaki” and “Hiroshima's specialty is oysters” as a result of inference, and can generate a response message.

<Process flow>
Next, a flow of processing executed by the robot 1 of the present embodiment will be described with reference to FIG. FIG. 7 is a flowchart illustrating an example of the flow of a process performed by the robot 1 according to the present embodiment. In the following description, it is assumed that the input content of the user includes the first word and the first related word, and the response message output by the robot 1 immediately before the input by the user includes the third word. Here, when the input content of the user is “Osaka's specialty is okonomiyaki”, the first word and the first related word are “okonomiyaki” and “specialty”, respectively, and “Osaka” is the first auxiliary word. is there.

  The processing in S11 to S22 is the same as in the first embodiment. In the case of YES in S22, the response message generator 26 generates a response message from the first phrase, the first accompanying phrase, and the second related phrase (S23), and the process proceeds to S27. Specifically, the response message generation unit 26 generates a response message “Osaka souvenir is Okonomiyaki” or “Hiroshima's specialty is oyster” in which the first word, the first accompanying word, and the second related word are associated with each other. . When the second phrase is not found in the processing of S20 (NO in S20), or when the second related phrase is not found in the processing of S22 (NO in S22), the response message generator 26 sets the first related phrase. A response message is generated from the word and the template (S24). Thereafter, the process proceeds to S27. If it is determined in S14 that the corresponding scenario information has been found (YES in S14), the response message generation unit 26 generates a response message from the first phrase, the first accompanying phrase, and the scenario information (S25), and performs processing. Goes to S27. Then, when it is determined in S16 that the concept of the first word and the third word are not the same (NO in S16), the response message generation unit 26 generates a response message from the first word, the first accompanying word, and the template. (S26), and the process proceeds to S27. The process of S27 is the same as that of the first embodiment.

  With the above-described processing, the robot 1 according to the present embodiment can perform the processing based on the words included in the input content even if a conversation scenario that matches the case where the input content of the user includes three or more words is not found. A response message can be generated.

(Modification)
In the above embodiments, the configuration in which the robot 1 interacts with the user by voice has been described, but the present invention is not limited to this. For example, the control unit 20 may be provided in a response device that interacts with the user using text. Further, the dialogue between the user and the response device by voice data or text may be performed via a network.

[Example of software implementation]
The control unit 20 (especially the response message generation unit 26 and the concept determination unit 24) of the robot 1 may be realized by a logic circuit (hardware) formed on an integrated circuit (IC chip) or the like, or a CPU (Central Processing). Unit) may be realized by software.

  In the latter case, the control unit 20 includes a CPU that executes instructions of a program that is software for realizing each function, a ROM (Read Only Memory) in which the program and various data are recorded so as to be readable by a computer (or CPU), or A storage device (these are referred to as “recording media”), a RAM (Random Access Memory) for expanding the above program, and the like are provided. Then, the object of the present invention is achieved when the computer (or CPU) reads the program from the recording medium and executes the program. As the recording medium, a “temporary tangible medium”, for example, a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. Further, the program may be supplied to the computer via an arbitrary transmission medium (a communication network, a broadcast wave, or the like) capable of transmitting the program. Note that the present invention can also be realized in the form of a data signal embedded in a carrier wave, in which the program is embodied by electronic transmission.

[Summary]
A response device (1) according to an aspect 1 of the present invention is a response device (1) that outputs a response message about an input content input by a user, wherein a first word and a first related word are related to each other. A storage unit (13) that stores that the second phrase and the second related phrase are related to each other, and a response message generation unit (26) that generates a response message to the input content. The first word or the second word is included in the input content, and the storage unit (13) determines that the second word and the first related word are related to each other, or In a case where the fact that the first phrase and the second phrase belong to the same concept is stored, the response message generation unit (26) performs a process of associating the first phrase and the second related phrase with each other. Response message above Generated.

  According to the above configuration, the response device includes a first related phrase and a first related phrase that are mutually related to the first or second phrase included in the input content input by the user and the first phrase stored in the storage unit. That the second word and the first word are related to each other, or that the first word and the second word belong to the same concept based on the second word related to the second word; Is stored, it is possible to generate a response message having a content in which the first word and the second related word are associated with each other. Therefore, even when a suitable conversation scenario cannot be found, there is an effect that a response device that generates a response message based on a phrase included in the input content is realized.

  In the response device (1) according to the second aspect of the present invention, in the first aspect, the storage unit (13) stores the relationship between the second phrase and the second related phrase in the relationship. The response message generation unit (26) generates the response message having contents in which the first phrase and the second related phrase are associated with each other by the relevance.

  According to the above configuration, the response device generates a response message having contents in which the first word and the second word are associated with each other, based on the relevance between the second word and the second word. Becomes possible. Therefore, it is possible to realize a response device that generates a response message based on the phrase included in the input content and the relevance between the second phrase and the second related phrase even when a suitable conversation scenario is not found. It works.

  According to a third aspect of the present invention, in the response device according to the first or second aspect, the storage unit (13) may be configured such that the second phrase and the first related phrase are related to each other. Is stored, the response message generation unit (26) generates the response message having a content in which the first word and the second related word are associated with each other.

  According to the above configuration, when the storage unit stores that the second word and the first related word are related to each other, the response device associates the first word and the second related word with each other. It is possible to generate a response message with the content. Therefore, even if a suitable conversation scenario is not found, the content of the first word and the second word related to each other is based on the fact that the second word and the first word are related to each other. This has the effect of realizing a response device that generates a response message.

  A response device (1) according to an aspect 4 of the present invention is the response device according to the aspect 1 or 2, wherein the storage unit (13) stores that the first word and the second word belong to the same concept. If so, the response message generation unit (26) generates the response message having contents in which the first word and the second related word are associated with each other.

  According to the above configuration, when the storage unit stores that the first word and the second word belong to the same concept, the response device has a content in which the first word and the second related word are associated with each other. A response message can be generated. Therefore, even when a suitable conversation scenario is not found, a response message having contents in which the first word and the second related word are associated with each other based on the fact that the first word and the second word belong to the same concept. Is realized.

  A response device (1) according to an aspect 5 of the present invention, in any of the aspects 1 to 4, wherein the storage unit (13) determines that the second word and a plurality of related words are related to each other. If it is stored, the response message generating unit (26) preferentially extracts, as the second related phrase, a related phrase belonging to the same concept as the first related phrase among the plurality of related phrases, and The response message having a content in which one phrase and the second related phrase are associated with each other is generated.

  According to the above configuration, when the storage unit stores that the second word and the plurality of related words are related to each other, the response device has the same concept as the first related word among the plurality of related words. Is extracted preferentially as a second related phrase, and a response message having contents in which the first word and the second related word are associated with each other is produced. Therefore, even if a suitable conversation scenario is not found, when the fact that the second phrase and the plurality of related phrases are related to each other is stored, the same as the first related phrase among the plurality of related phrases is stored. This has the effect of realizing a response device that preferentially extracts a related phrase belonging to the concept as a second related phrase and generates a response message having a content in which the first word and the second related phrase are associated with each other.

  The response device (1) according to an aspect 6 of the present invention, according to any one of the aspects 1 to 5, wherein the output unit (12) for outputting the response message to the outside and the output unit immediately before the input content is input. A concept determining unit (24) that determines whether the concept to which the third phrase included in the response message output from the unit (12) belongs and the concept to which the input phrase included in the input content belongs are the same; The storage unit (13) stores the template of the response message, and the response message generation unit (26) determines that the concept to which the third word belongs and the concept to which the input word belongs are the same. In some cases, the response message is generated having a content in which the first word and the second related word are associated with each other. If the concept to which the third word belongs and the concept to which the input word belongs are different, the template It generates another response message using preparative.

  According to the above configuration, the response device includes a concept to which the third phrase included in the response message output immediately before input of the input content by the response device belongs, and a concept to which the input phrase included in the input content belongs. Are the same, a response message is generated having a content in which the first word and the second related word are associated with each other. If the concept to which the third word belongs and the concept to which the input word belongs are different, a response message Can be generated. Therefore, even when a suitable conversation scenario cannot be found, an effect is achieved that a response device that generates a response message based on the concept to which the third word belongs and the concept to which the input word belongs is provided.

  The response device (1) according to aspect 7 of the present invention, in the aspect 6, when the concept to which the third word belongs and the concept to which the input word belongs are different, the response message generation unit (26) A phrase is combined with a template associated with the third phrase to generate the another response message.

  According to the above configuration, when the concept to which the third word belongs and the concept to which the input word belongs are different, the response device combines the input word with the template related to the third word to generate another response message. It becomes possible. Therefore, even if a suitable conversation scenario is not found, if the concept to which the third word belongs and the concept to which the input word belongs are different, the input word is combined with a template related to the third word to obtain another response. This has the effect of realizing a response device that generates a message.

  The response device (1) according to aspect 8 of the present invention, in the aspect 6, when the concept to which the third word belongs and the concept to which the input word belongs differ, the response message generation unit (26) A word and a fourth word included in the input content are combined with a template to generate another response message.

  According to the configuration described above, when the concept to which the third word belongs and the concept to which the input word belongs are different, the response device combines the input word and the fourth word included in the input content with the template to generate another A response message can be generated. Therefore, even if a suitable conversation scenario is not found, if the concept to which the third word belongs and the concept to which the input word belongs are different, the input word and the fourth word included in the input content are included in the template. This has the effect of realizing a response device that generates another response message in combination.

  In the response device (1) according to aspect 9 of the present invention, in the aspect 7 or 8, the response message generation unit (26) includes, for a plurality of phrases included in the input content, a concept to which each phrase belongs and each phrase. The input word is selected based on the context information including the words before and after.

  According to the above configuration, the response device can select an input word based on the attribute to which each word belongs and context information including words before and after each word. Therefore, even if a suitable conversation scenario is not found, an input word is selected from attributes to which each word belongs and context information including words before and after each word and combined with a template related to the third word to obtain another word. This has the effect of realizing a response device that generates a response message.

  A control method for a response device (1) according to an aspect 10 of the present invention is a control method for a response device (1) that outputs a response message based on an input content input by a user, wherein the response device (1) includes a A storage unit for storing that one word and the first related word are related to each other and that the second word and the second related word are related to each other; A search step (S17) for searching the storage unit for the storage of the first word or the second word, and a response message generation step (S23) of generating a response message to the input content. Alternatively, the second word is included in the input content, and the storage unit stores information that the second word and the first related word are related to each other, or that the first word and the second word are related to each other. If is that stores that belong to the same concept, in the response message generation step (S23), it generates a response message content associated with each other and the first word and the second related phrases. According to the above configuration, the same operation and effect as those of the first aspect are exerted.

  The response device (1) according to each aspect of the present invention may be realized by a computer. In this case, the response device is operated by operating a computer as each unit (software element) included in the response device (1). A control program of the response device (1) for realizing (1) by a computer and a computer-readable recording medium recording the control program are also included in the scope of the present invention.

  The present invention is not limited to the embodiments described above, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention. Furthermore, new technical features can be formed by combining the technical means disclosed in each embodiment.

Reference Signs List 1 robot (response device) 11 voice input unit 12 voice output unit (output unit)
Reference Signs List 13 storage unit 20 control unit 21 speech recognition unit 22 word / concept acquisition unit 23 scenario unit 24 concept determination unit 25 related phrase acquisition unit 26 response message generation unit 27 speech synthesis unit

Claims (11)

A response device that outputs a response message about an input content input by a user,
A storage unit that stores that the first word and the first related word are related to each other, and that the second word and the second related word are related to each other;
A response message generator that generates a response message to the input content,
The first word or the second word is included in the input content, and the storage unit determines that the second word and the first related word are related to each other, or that the first word or the second word is related to the first word. If you remember that the second phrase belongs to the same concept,
The response device, wherein the response message generation unit generates the response message having a content in which the first phrase and the second related phrase are associated with each other. The storage unit stores how the second word and the second related word are related to each other,
The response device according to claim 1, wherein the response message generation unit generates the response message having a content in which the first word and the second related word are associated with each other by the association. When the storage unit stores that the second phrase and the first related phrase are related to each other,
The response device according to claim 1, wherein the response message generation unit generates the response message having a content in which the first word and the second related word are associated with each other. When the storage unit stores that the first phrase and the second phrase belong to the same concept,
The response device according to claim 1, wherein the response message generation unit generates the response message having a content in which the first word and the second related word are associated with each other. When the storage unit stores that the second word and a plurality of related words are related to each other,
The response message generating unit preferentially extracts a related term belonging to the same concept as the first related term from the plurality of related terms as the second related term, and extracts the first term and the second related term. The response device according to any one of claims 1 to 4, wherein the response message having a content in which the response messages are associated with each other is generated. An output unit that outputs the response message to the outside,
It is determined whether or not the concept to which the third phrase included in the response message output from the output unit immediately before the input content belongs is the same as the concept to which the input phrase included in the input content belongs. And a concept determination unit that performs
The storage unit stores a template of the response message,
The response message generator,
When the concept to which the third word belongs and the concept to which the input word belongs are the same, the above-mentioned response message is generated with the first word and the second related word being associated with each other;
The response device according to any one of claims 1 to 5, wherein when a concept to which the third word belongs and a concept to which the input word belongs are different, another response message is generated using a template. .   If the concept to which the third word belongs and the concept to which the input word belongs are different, the response message generation unit generates the another response message by combining the input word with a template related to the third word. 7. The response device according to claim 6, wherein:   When the concept to which the third word belongs and the concept to which the input word belongs are different, the response message generation unit combines the input word and the fourth word included in the input content with a template to generate the another word. The response device according to claim 6, wherein the response device generates a response message.   The said response message generation part selects the said input phrase based on the context information containing the concept to which each phrase belongs, and the phrase before and after each phrase about several phrases contained in the said input content, The Claims characterized by the above-mentioned. 9. The response device according to 7 or 8. A control method of a response device that outputs a response message for an input content input by a user,
The response device includes a storage unit that stores that the first word and the first related word are related to each other, and that the second word and the second related word are related to each other,
A search step of searching the storage unit for storage relating to the first phrase or the second phrase included in the input content;
A response message generation step of generating a response message to the input content,
The first word or the second word is included in the input content, and the storage unit stores information that the second word and the first related word are related to each other, or that the first word and the second word are related to each other. If you remember that the second phrase belongs to the same concept,
In the response message generating step, a control method of a response device, comprising generating a response message having a content in which the first phrase and the second related phrase are associated with each other.   A control program for causing a computer to function as the response device according to claim 1, wherein the control program causes the computer to function as the response message generation unit.

Images