JP2013072887A - Interactive device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve a natural interaction.SOLUTION: An interaction device includes an utterance set storage unit, an utterance set acquisition unit, a first output unit, a detection unit, and a second output unit. In the utterance set storage unit, an utterance set including a first utterance and a second utterance representing an utterance of a response to user's utterance assumed as a response to the first utterance is stored. The utterance set acquisition unit acquires an utterance set. The first output unit outputs the first utterance included in the acquired utterance set. The detection unit detects user's utterance after output of the first utterance. On detection of user's utterance, the second output unit outputs the second utterance included in the acquired utterance set.

Description

  Embodiments described herein relate generally to an interactive apparatus.

  Conventionally, an interactive apparatus that interacts with a user is known. The dialogue using the dialogue device includes, for example, a dialogue that grasps a user's input sentence and a dialogue that grasps a user's voice. In such an interactive apparatus, there are cases where advanced language processing and speech recognition processing are performed in order to accurately grasp a user's input sentence or voice and realize a more natural conversation.

JP 2006-331343 A JP 2006-31467 A JP 2010-79574 A

  However, in the prior art, there is a problem that it is difficult to realize a natural dialogue when the user's input sentence or voice cannot be accurately grasped. Specifically, in the prior art, errors may occur in language processing and speech recognition processing, so it is difficult to realize natural conversation.

  The problem to be solved by the present invention is to provide an interactive apparatus capable of realizing natural conversation.

  The interactive apparatus according to the embodiment includes an utterance set storage unit, an utterance set acquisition unit, a first output unit, a detection unit, and a second output unit. The utterance set storage unit stores an utterance set including a first utterance and a second utterance representing an utterance of a response to an utterance by a user assumed as a response to the first utterance. The utterance set acquisition unit acquires an utterance set. The first output unit outputs the first utterance included in the acquired utterance set. The detection unit detects an utterance by the user after the first utterance is output. The second output unit outputs the second utterance included in the acquired utterance set when the utterance by the user is detected.

FIG. 1 is a block diagram illustrating a configuration example of the interactive apparatus according to the first embodiment. FIG. 2 is a diagram illustrating an example of information stored in the utterance set storage unit according to the first embodiment. FIG. 3 is a flowchart illustrating an example of the flow of the interactive processing according to the first embodiment. FIG. 4 is a block diagram illustrating a configuration example of the interactive apparatus according to the second embodiment. FIG. 5A is a diagram illustrating an example of information stored in the utterance template storage unit according to the second embodiment. FIG. 5B is a diagram illustrating an example of information stored in the utterance template storage unit according to the second embodiment. FIG. 6 is a flowchart illustrating an example of the flow of an utterance set generation process according to the second embodiment. FIG. 7 is a block diagram illustrating a configuration example of the interactive apparatus according to the third embodiment. FIG. 8 is a diagram illustrating an example of information stored in the specific word storage unit according to the third embodiment. FIG. 9 is a flowchart illustrating an example of the flow of an utterance set generation process according to the third embodiment. FIG. 10 is a block diagram illustrating a configuration example of the interactive apparatus according to the fourth embodiment. FIG. 11 is a flowchart illustrating an example of the flow of an utterance set generation process according to the fourth embodiment. FIG. 12 is a block diagram illustrating a configuration example of the interactive apparatus according to the fifth embodiment. FIG. 13 is a diagram illustrating an example of information stored in the utterance template storage unit according to the fifth embodiment. FIG. 14 is a diagram for explaining the order determination process according to the fifth embodiment. FIG. 15 is a flowchart illustrating an example of a flow of interactive processing according to the fifth embodiment. FIG. 16 is a block diagram illustrating a configuration example of the interactive apparatus according to the sixth embodiment. FIG. 17 is a flowchart illustrating an example of the flow of interactive processing according to the sixth embodiment. FIG. 18 is a block diagram illustrating a configuration example of the interactive apparatus according to the seventh embodiment. FIG. 19 is a flowchart illustrating an example of the flow of interactive processing according to the seventh embodiment. FIG. 20 is a flowchart illustrating an example of the flow of dialogue processing related to a request for recurrent speech. FIG. 21 is a diagram illustrating an example of information stored in the utterance set storage unit including the third utterance. FIG. 22 is a flowchart illustrating an example of the flow of interactive processing when a user utterance cannot be detected at a certain time.

(First embodiment)
FIG. 1 is a block diagram illustrating a configuration example of the interactive apparatus according to the first embodiment. For example, as illustrated in FIG. 1, the dialogue apparatus 100 includes an utterance set storage unit 101, an utterance set acquisition unit 102, an output unit 103, and a detection unit 104. For example, the interactive device 100 realizes an interaction with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The utterance set storage unit 101 stores an utterance set including a first utterance and a second utterance representing an utterance of a response to an utterance by a user assumed as a response to the first utterance. Here, information stored in the utterance set storage unit 101 according to the first embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of information stored in the utterance set storage unit 101 according to the first embodiment.

  For example, as illustrated in FIG. 2, the utterance set storage unit 101 includes a first utterance “Is a movie?” And an assumed user utterance “I like” representing an utterance by a user assumed for the first utterance, The “utterance set 1” including the second utterance “Yes” is stored. In FIG. 2, for convenience of explanation, the utterance set storage unit 101 including “assumed user utterance” is illustrated, but the “assumed user utterance” may not be included in the utterance set storage unit 101. That is, as shown in FIG. 2, the utterance set storage unit 101 stores an utterance set that can realize a natural conversation regardless of the response of the user utterance to the first utterance.

  The utterance set acquisition unit 102 acquires an utterance set. For example, the utterance set acquisition unit 102 stores the utterance set stored in the utterance set storage unit 101 at an arbitrary timing such as when the user stands in front of the dialog device 100 or when the user is ready to use the dialog device 100. To get. For example, the utterance set acquisition unit 102 acquires the first utterance “Is a movie?” And the second utterance “I think so” that are “utterance set 1” stored in the utterance set storage unit 101. To do.

  The output unit 103 outputs the utterance set acquired by the utterance set acquisition unit 102. The output unit 103 includes a first output unit 103a and a second output unit 103b. Among these, the first output unit 103 a outputs the first utterance included in the utterance set acquired by the utterance set acquisition unit 102. For example, the first output unit 103a outputs the first utterance “Do you like movies?” Included in the utterance set “utterance set 1” acquired by the utterance set acquisition unit 102 to a predetermined display output device. The processing by the second output unit 103b will be described later.

  The detection unit 104 detects an utterance by the user after the first utterance is output. For example, the detection unit 104 detects an utterance by the user after the first output unit 103a outputs the first utterance. For example, the detection unit 104 detects that the user uttered “I like” after the first utterance “Is movie?” Is output to a predetermined display output device by the first output unit 103a. To do. Note that the utterance by the user is not limited to the above example, and may be any utterance.

  The second output unit 103 b outputs the second utterance included in the utterance set acquired by the utterance set acquisition unit 102 when the detection unit 104 detects an utterance by the user. For example, the second output unit 103b includes the second output unit 103b included in the utterance set “utterance set 1” acquired by the utterance set acquisition unit 102 when the detection unit 104 detects an utterance “I like” by the user. Two utterances “Yes” are output to a predetermined display output device. However, if the detection unit 104 cannot detect the user's utterance for a certain period of time, the utterance set acquisition unit 102 acquires the next utterance set without outputting the second utterance by the second output unit 103b.

  Next, the flow of interactive processing according to the first embodiment will be described with reference to FIG. FIG. 3 is a flowchart illustrating an example of the flow of the interactive processing according to the first embodiment.

  For example, as illustrated in FIG. 3, the utterance set acquisition unit 102 acquires an utterance set from the utterance set storage unit 101 (step S101). The first output unit 103a outputs the first utterance included in the utterance set acquired by the utterance set acquisition unit 102 to a predetermined display output device (Step S102).

  Moreover, the detection part 104 determines whether the user's utterance with respect to the 1st utterance output by the 1st output part 103a was detected (step S103). At this time, when the user's utterance is detected by the detection unit 104 (Yes in step S103), the second output unit 103b displays a predetermined display of the second utterance included in the utterance set acquired by the utterance set acquisition unit 102. Output to the output device (step S104). After the output of the second utterance, the utterance set acquisition unit 102 acquires the next utterance set from the utterance set storage unit 101 (step S101).

  On the other hand, when the utterance by the user is not detected (No at Step S103), the detection unit 104 determines whether or not a predetermined time has elapsed (Step S105). At this time, when the predetermined time has not elapsed (No at Step S105), the detection unit 104 determines again whether or not the user's utterance has been detected (Step S103). On the other hand, when the predetermined time has elapsed (Yes at Step S105), the utterance set acquisition unit 102 acquires the next utterance set from the utterance set storage unit 101 (Step S101).

  According to the present embodiment, since the dialogue is performed using the utterance that limits the utterance by the user and the utterance of the response to the utterance by the user, a natural dialogue can be realized. In other words, a natural conversation can be realized without recognizing the content of the utterance by the user.

(Second Embodiment)
FIG. 4 is a block diagram illustrating a configuration example of the interactive apparatus according to the second embodiment. In the second embodiment, functional units that perform the same processes as in the first embodiment are denoted by the same reference numerals, and descriptions of the same processes may be omitted.

  For example, as illustrated in FIG. 4, the dialogue apparatus 200 includes an utterance set storage unit 101, an utterance set acquisition unit 102, an output unit 103, a detection unit 104, a concept dictionary storage unit 205, and an utterance template storage unit 206. A concept acquisition unit 207, an utterance template acquisition unit 208, and an utterance set generation unit 209. Similarly to the first embodiment, the dialogue apparatus 200 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The concept dictionary storage unit 205 stores words and concepts in association with each other. For example, the words stored in the concept dictionary storage unit 205 are commonly used words and do not include proper nouns, new words, and the like. The utterance template storage unit 206 stores an utterance set template including a first utterance and a second utterance in which some words are expressed in concept. Here, information stored in the utterance template storage unit 206 according to the second embodiment will be described with reference to FIGS. 5A and 5B. 5A and 5B are diagrams illustrating examples of information stored in the utterance template storage unit 206 according to the second embodiment.

  For example, as illustrated in FIG. 5A, the utterance template storage unit 206 has a first utterance “Do you like [food]?”, An assumed user utterance “I like you”, and the second utterance “[food] is delicious. The template “template 1” of the utterance set including “yo” is stored. Here, “food” surrounded by “[]” is a concept of a word. That is, in the utterance set template stored in the utterance template storage unit 206, words included in the utterance are expressed in concept.

  For example, as illustrated in FIG. 5B, the utterance template storage unit 206 includes a first utterance “Do you like [food]?”, An assumed user utterance “I like you”, and a second utterance “[food]” The template “template 1” of the utterance set including “taste” is stored. Similarly, “food” and “taste” surrounded by “[]” are word concepts. That is, in the utterance set template stored in the utterance template storage unit 206, a plurality of words included in the utterance may be expressed in concept. 5A and 5B illustrate the utterance template storage unit 206 including “assumed user utterance” for convenience of explanation, but the “assumed user utterance” may not be included in the utterance template storage unit 206. .

  When a phrase is input, the concept acquisition unit 207 acquires a concept corresponding to the phrase from the concept dictionary storage unit 205. For example, the concept acquisition unit 207 receives an input of the word “chocolate” by the operation of the interactive device 200 by the user. Then, the concept acquisition unit 207 acquires the concept “food” corresponding to the received word “chocolate” from the concept dictionary storage unit 205. Further, as an example in the case where a plurality of words are input, the concept acquisition unit 207 accepts input of the words “chocolate” and “sweet” by the operation of the interactive device 200 by the user. Then, the concept acquisition unit 207 acquires the concepts “food” and “taste” corresponding to the received words “chocolate” and “sweet” from the concept dictionary storage unit 205. Note that the dialogue apparatus 200 ends the process when the concept corresponding to the word is not registered in the concept dictionary storage unit 205.

  The utterance template acquisition unit 208 acquires an utterance set template including the concept acquired by the concept acquisition unit 207 from the utterance template storage unit 206. For example, the utterance template acquisition unit 208 acquires the template “template 1” of the utterance set including the concept “food” acquired by the concept acquisition unit 207 from the utterance template storage unit 206 (see FIG. 5A). Further, as an example in the case where a plurality of words are input, the utterance template acquisition unit 208 uses the template “template 1” of the utterance set including the concepts “food” and “taste” acquired by the concept acquisition unit 207. Acquired from the utterance template storage unit 206 (see FIG. 5B). Note that the dialogue apparatus 200 ends the process when the template of the utterance set including the concept is not registered in the utterance template storage unit 206.

  The utterance set generation unit 209 generates a new utterance set by inserting the input phrase into the concept included in the utterance set template acquired by the utterance template acquisition unit 208. Then, the utterance set generation unit 209 stores the generated new utterance set in the utterance set storage unit 101.

  For example, the utterance set generation unit 209 receives the word “chocolate” received from the concept “food” included in the template “template 1” (see FIG. 5A) of the utterance set acquired by the utterance template acquisition unit 208. ”To generate a new utterance set. The utterance set generated in the above example includes the first utterance “Do you like chocolate?”, The assumed user utterance “I like you”, and the second utterance “I love chocolate”. Then, the utterance set generation unit 209 stores the generated new utterance set in the utterance set storage unit 101.

  Further, when an example in which a plurality of words are input is given, the utterance set generation unit 209 includes a concept “template 1” (see FIG. 5B) included in the utterance set acquired by the utterance template acquisition unit 208. A new utterance set is generated by inserting the input words “chocolate” and “sweet” into “food” and “taste”. The utterance set generated in the above example includes the first utterance “Do you like chocolate?”, The assumed user utterance “I like you”, and the second utterance “Chocolate is sweet”. Then, the utterance set generation unit 209 stores the generated new utterance set in the utterance set storage unit 101.

  Next, the flow of the utterance set generation process according to the second embodiment will be described with reference to FIG. FIG. 6 is a flowchart illustrating an example of the flow of an utterance set generation process according to the second embodiment.

  For example, as shown in FIG. 6, when the concept acquisition unit 207 receives an input of a word (Yes at Step S201), the concept acquisition unit 207 searches the concept dictionary storage unit 205 for a concept corresponding to the word (Step S202). In addition, when the concept acquisition unit 207 does not accept an input of a word (No at Step S201), the concept acquisition unit 207 enters a state of waiting for an input of the word.

  At this time, when the concept corresponding to the word is acquired from the concept dictionary storage unit 205 by the concept acquisition unit 207 (Yes in step S203), the utterance template acquisition unit 208 utters the template of the utterance set including the acquired concept. A search is made from the template storage unit 206 (step S204). On the other hand, when the concept corresponding to the word is not acquired from the concept dictionary storage unit 205 by the concept acquisition unit 207 (No in step S203), the interactive device 200 ends the process.

  Also, the utterance template acquisition unit 208 searches the utterance template storage unit 206 for an utterance set template including the concept acquired by the concept acquisition unit 207 (step S204). At this time, when the utterance set template is acquired from the utterance template storage unit 206 by the utterance template acquisition unit 208 (Yes in step S205), the utterance set generation unit 209 uses the concept included in the acquired utterance set template. Then, the input word is inserted to generate a new utterance set (step S206). Thereafter, the utterance set generation unit 209 stores the generated utterance set in the utterance set storage unit 101. On the other hand, when the utterance template acquisition unit 208 has not acquired the utterance set template from the utterance template storage unit 206 (No at Step S205), the interactive apparatus 200 ends the process.

  According to the present embodiment, not only a dialogue using a previously registered utterance set but also various dialogues using a utterance set based on an arbitrary word can be realized.

(Third embodiment)
FIG. 7 is a block diagram illustrating a configuration example of the interactive apparatus according to the third embodiment. In the third embodiment, functional units that perform the same processes as those in the first or second embodiment are denoted by the same reference numerals, and the description of the same processes may be omitted.

  For example, as shown in FIG. 7, the dialogue apparatus 300 includes an utterance set storage unit 101, an utterance set acquisition unit 102, an output unit 103, a detection unit 104, an utterance template storage unit 206, and an utterance template acquisition unit 308. And an utterance set generation unit 309 and a unique word storage unit 310. Similarly to the first embodiment, the dialogue apparatus 300 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The unique word storage unit 310 stores a unique phrase acquired from the outside and the concept of the unique phrase in association with each other. Here, information stored in the unique word storage unit 310 according to the third embodiment will be described with reference to FIG. FIG. 8 is a diagram illustrating an example of information stored in the unique word storage unit 310 according to the third embodiment.

  For example, as illustrated in FIG. 8, the unique word storage unit 310 includes a unique word “hot”, a unique word concept “weather”, and a feature “http://tenki_jouhou.jp” representing the origin of the unique word. Store in association with each other. The unique word is an example of a unique phrase. Information stored in the unique word storage unit 310 is updated as appropriate based on information that changes daily. Specifically, the specific word corresponding to the weather is updated to a word (for example, “sunny”, “rain”, “hot weather”, “guerrilla heavy rain”, etc.) based on the weather registered in advance. Similarly, the unique word corresponding to the news is updated to a word (for example, “Nadeshiko Japan”, “Great East Japan Earthquake”, “OO Prime Minister”, etc.) based on the analyzed article on the specific news site. Similarly, the other unique words are updated to words that frequently appear in micro blogging (for example, “Typhoon XX”, “Sincerity Sincerity”, “Girls Association”, etc.). As described above, the unique words stored in the unique word storage unit 310 are words having different properties from the words registered in the concept dictionary storage unit 205 and the like described above, including words that are not used regularly. The registration of the unique word in the unique word storage unit 310 may be executed in the interactive apparatus 300, or a service for extracting the unique word may be used.

  The utterance template acquisition unit 308 acquires an utterance set template including a unique phrase concept from the utterance template storage unit 206. For example, the utterance template acquisition unit 308 acquires the concept “food” of the unique word “wasabi software” stored in the unique word storage unit 310. Then, the utterance template acquisition unit 308 acquires the template “template 1” of the utterance set including the acquired concept “food” from the utterance template storage unit 206 (see FIG. 5A). Note that when the utterance template acquisition unit 308 acquires the concept of the specific word from the specific word storage unit 310, as one aspect, the utterance template acquisition unit 308 acquires the registration in the specific word storage unit 310 from the latest one.

  Further, as an example in the case of acquiring the concept of a plurality of unique words, the utterance template acquisition unit 308 has a concept of the unique words “wasabi soft” and “spicy enough to tear” stored in the unique word storage unit 310. Get “food” and “taste”. Then, the utterance template acquisition unit 308 acquires the template “template 1” of the utterance set including the acquired concepts “food” and “taste” from the utterance template storage unit 206 (see FIG. 5B).

  The utterance set generation unit 309 generates a new utterance set by inserting a unique word / phrase into the concept included in the utterance set template acquired by the utterance template acquisition unit 308. Then, the utterance set generation unit 309 stores the generated new utterance set in the utterance set storage unit 101.

  For example, the utterance set generation unit 309 adds the unique word “wasabi software” to the concept “food” included in the template “template 1” (see FIG. 5A) of the utterance set acquired by the utterance template acquisition unit 308. Insert a new utterance set. The utterance set generated in the above example includes the first utterance “Do you like wasabi software?”, The assumed user utterance “I like it”, and the second utterance “I like wasabi software”. Become. Then, the utterance set generation unit 309 stores the generated new utterance set in the utterance set storage unit 101.

  Further, when an example in which a concept of a plurality of unique words is acquired, the utterance set generation unit 309 is included in the utterance set template “template 1” (see FIG. 5B) acquired by the utterance template acquisition unit 308. A new utterance set is generated by inserting the unique words “wasabi soft” and “spicy enough to tear” into the concepts “food” and “taste”. The utterance set generated in the above example is the first utterance “Do you like wasabi software?”, The assumed user utterance “I like it”, and the second utterance “wasabi software is so hot that tears come out” Will be included. Then, the utterance set generation unit 309 stores the generated new utterance set in the utterance set storage unit 101.

  In addition, when acquiring the concept of a some specific word from the specific word memory | storage part 310, it is preferable to acquire the concept of the specific word which has the same feature. This is because a more suitable utterance can be generated by generating an utterance using specific words having some relationship. For this reason, in the above example, the unique words “wasabi soft” and “food” and “taste” of “wasabi soft” having the same feature “http: //web.Analyze.cgi” are acquired. The case is given as an example.

  Next, the flow of an utterance set generation process according to the third embodiment will be described with reference to FIG. FIG. 9 is a flowchart illustrating an example of the flow of an utterance set generation process according to the third embodiment.

  For example, as shown in FIG. 9, the utterance template acquisition unit 308 acquires the concept of the specific word from the specific word storage unit 310 (step S301). Then, the utterance template acquisition unit 308 searches the utterance template storage unit 206 for an utterance set template including the acquired concept of the specific word (step S302).

  At this time, when the utterance template acquisition unit 308 acquires the template of the corresponding utterance set from the utterance template storage unit 206 (Yes in step S303), the utterance set generation unit 309 is included in the acquired utterance set template. A new utterance set is generated by inserting the acquired unique word into the concept (step S304).

  Thereafter, the utterance set generation unit 309 stores the generated utterance set in the utterance set storage unit 101. On the other hand, when the utterance template acquisition unit 308 does not acquire the utterance set template from the utterance template storage unit 206 (No at Step S303), the interactive apparatus 300 ends the process.

  According to the present embodiment, it is possible to realize a dialogue using an utterance set including the latest keyword updated daily.

(Fourth embodiment)
FIG. 10 is a block diagram illustrating a configuration example of the interactive apparatus according to the fourth embodiment. In the fourth embodiment, functional units that perform the same processes as those in the first embodiment or the second embodiment are denoted by the same reference numerals, and descriptions of the same processes may be omitted.

  For example, as shown in FIG. 10, the dialogue apparatus 400 includes an utterance set storage unit 101, an utterance set acquisition unit 102, an output unit 103, a detection unit 104, a concept dictionary storage unit 205, and an utterance template storage unit 206. A concept acquisition unit 407, an utterance template acquisition unit 208, an utterance set generation unit 409, and an extraction unit 411. Similarly to the first embodiment, the dialogue apparatus 400 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The extraction unit 411 extracts words / phrases included in the user's utterance that match the words / phrases stored in the concept dictionary storage unit 205. For example, the extraction unit 411 performs speech recognition of an utterance made by the user detected by the detection unit 104. Then, the extraction unit 411 extracts words remaining as a result of the speech recognition processing as words included in the utterance by the user that match the words stored in the concept dictionary storage unit 205.

  In such a speech recognition process, as one aspect, a utterance by a user is divided into words by performing a morphological analysis, and a process of searching each word from the concept dictionary storage unit 205 is executed. Here, words that are not stored in the concept dictionary storage unit 205 are excluded because of particles, auxiliary verbs, and the like. As a result, the remaining words are extracted as a result of the speech recognition process. Here, a case where the word remaining as a result of the speech recognition process is “curry” is taken as an example.

  The concept acquisition unit 407 acquires a concept corresponding to the word extracted by the extraction unit 411 from the concept dictionary storage unit 205. For example, the concept acquisition unit 407 acquires the concept “food” corresponding to the remaining word “curry” from the concept dictionary storage unit 205 as a result of the speech recognition processing by the extraction unit 411. Similarly to the second embodiment, the utterance template acquisition unit 208 uses the utterance template storage unit 206 (see FIG. 5A) for the template “template 1” of the utterance set including the concept “food” acquired by the concept acquisition unit 407. )

  The utterance set generation unit 409 generates a new utterance set by inserting the phrase extracted by the extraction unit 411 into the concept included in the utterance set template acquired by the utterance template acquisition unit 208. Then, the utterance set generation unit 409 stores the generated new utterance set in the utterance set storage unit 101.

  For example, the utterance set generation unit 409 is extracted by the extraction unit 411 into the concept “food” included in the template “template 1” (see FIG. 5A) of the utterance set acquired by the utterance template acquisition unit 208. Insert the word “curry” to generate a new utterance set. The utterance set generated in the above example includes the first utterance “Do you like curry?”, The assumed user utterance “I like you”, and the second utterance “I like curry is delicious”. Then, the utterance set generation unit 409 stores the generated new utterance set in the utterance set storage unit 101.

  Next, the flow of an utterance set generation process according to the fourth embodiment will be described with reference to FIG. FIG. 11 is a flowchart illustrating an example of the flow of an utterance set generation process according to the fourth embodiment.

  For example, as shown in FIG. 11, when the utterance by the user is detected by the detection unit 104 (Yes in step S401), the extraction unit 411 executes a speech recognition process to thereby extract a word included in the utterance by the user. Extract (step S402). In addition, when the detection unit 104 has not detected the user's utterance (No in step S401), the extraction unit 411 enters a state of waiting for detection of the user's utterance by the detection unit 104.

  The concept acquisition unit 407 searches the concept dictionary storage unit 205 for a concept corresponding to the word extracted by the extraction unit 411 (step S403). At this time, when the concept corresponding to the word is acquired from the concept dictionary storage unit 205 by the concept acquisition unit 407 (Yes in step S404), the utterance template acquisition unit 208 utters the template of the utterance set including the acquired concept. A search is made from the template storage unit 206 (step S405). On the other hand, when the concept corresponding to the word is not acquired from the concept dictionary storage unit 205 by the concept acquisition unit 407 (No at Step S404), the dialogue apparatus 400 ends the process.

  Further, when the utterance set template is acquired from the utterance template storage unit 206 by the utterance template acquisition unit 208 (Yes in step S406), the utterance set generation unit 409 adds the concept included in the acquired utterance set template to the concept. The words extracted by the extraction unit 411 are inserted to generate a new utterance set (step S407). Thereafter, the utterance set generation unit 409 stores the generated utterance set in the utterance set storage unit 101. On the other hand, when the utterance template acquisition unit 208 does not acquire the utterance set template from the utterance template storage unit 206 (No at step S406), the interactive apparatus 400 ends the process.

  According to the present embodiment, since the dialogue is performed using the utterance set generated based on the user utterance, a natural flow of dialogue can be realized.

(Fifth embodiment)
FIG. 12 is a block diagram illustrating a configuration example of the interactive apparatus according to the fifth embodiment. In the fifth embodiment, functional units that perform the same processes as those in the first embodiment or the second embodiment are denoted by the same reference numerals, and the description of the same processes may be omitted.

  For example, as shown in FIG. 12, the dialogue apparatus 500 includes an utterance set storage unit 101, an utterance set acquisition unit 502, an output unit 103, a detection unit 104, a concept dictionary storage unit 205, and an utterance template storage unit 506. A concept acquisition unit 207, an utterance template acquisition unit 208, an utterance set generation unit 209, a similarity calculation unit 512, a co-occurrence dictionary storage unit 513, and a determination unit 514. Similarly to the first embodiment, the dialogue apparatus 500 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  Similar to the second embodiment, the utterance template storage unit 506 stores a template of an utterance set including a first utterance and a second utterance in which some words are expressed as concepts. Furthermore, the utterance template storage unit 506 associates the type of the first utterance, the user information obtained from the utterance by the user with respect to the first utterance, and the topic of the template of the utterance set with the first utterance and the second utterance. Remember. Here, information stored in the utterance template storage unit 506 according to the fifth embodiment will be described with reference to FIG. FIG. 13 is a diagram illustrating an example of information stored in the utterance template storage unit 506 according to the fifth embodiment.

  For example, as illustrated in FIG. 13, the utterance template storage unit 506 includes a first utterance type “question”, a first utterance “Do you like [food]?”, And an assumed user utterance “I like you”. A template “template 1” of an utterance set including information “favorite food”, the second utterance “[food] is [taste]”, and the topic “food” is stored. Here, “food” and “taste” surrounded by “[]” are word concepts.

  In addition, the utterance template storage unit 506 recommends the first utterance type “offer”, the first utterance “How is [food]?”, The assumed user utterance “mo”, and the second utterance “[food]”. ”And a template“ template 7 ”of an utterance set including the topic“ food ”. Similarly, “food” surrounded by “[]” is a concept of a word. Like the “template 7”, the user information may not be obtained depending on the template of the utterance set. If the user information cannot be obtained, the information is not retained. Note that the “assumed user utterance” may not be included in the utterance template storage unit 506.

  The similarity calculation unit 512 calculates the similarity between utterance sets including a new utterance set or between templates of an utterance set. For example, the similarity calculation unit 512 calculates the similarity based on the “first utterance type” between the templates of the utterance set. Specifically, the similarity calculation unit 512 calculates the similarity “Rt (s1, s2)” based on the type of the first utterance between the utterance set template “s1” and the utterance set template “s2” as the first utterance. If the types are the same, “1” is set, and if they are different, “0” is set. For example, the similarity calculation unit 512 calculates the similarity “Rt (s1, s2)” based on the type of the first utterance between the template “template 1” of the utterance set and the template “template 6” of the utterance set. Since the type of the first utterance is the same “question”, “1” is set. Similarly, the similarity calculation unit 512 calculates the similarity “Rt (s1, s2)” based on the type of the first utterance between the template “template 1” of the utterance set and the template “template 7” of the utterance set as the first. Since the type of utterance is “question” and “offer”, “0” is set. The reason for calculating the similarity based on the first utterance is to prevent the conversations that are similar to each other from being continued by causing the same utterance types to be continued.

  Further, for example, the similarity calculation unit 512 calculates the similarity based on “user information” between the templates of the utterance set. Specifically, the similarity calculation unit 512 sets the similarity “Ru (s1, s2)” based on user information between the utterance set template “s1” and the utterance set template “s2” if the user information is the same. “1” if it is different, “0” if it is different. For example, the similarity calculation unit 512 indicates the similarity “Ru (s1, s2)” based on the user information between the utterance set template “template 1” and the utterance set template “template 6”. Since it is the same “favorite food”, “1” is set. Similarly, the similarity calculation unit 512 changes the similarity “Ru (s1, s2)” based on user information between the utterance set template “template 1” and the utterance set template “template 7” to different “favorite foods”. ”And“ (blank) ”, so“ 0 ”. The reason why the similarity based on the user information is calculated is that it is not preferable to continue conversations asking for similar things.

  Further, for example, the similarity calculation unit 512 calculates the similarity based on “topic” between the templates of the utterance set. Specifically, the similarity calculation unit 512 determines the similarity “Rd (s1, s2)” based on the topic between the utterance set template “s1” and the utterance set template “s2” if the topics are the same. “1”, otherwise “0”. For example, the similarity calculation unit 512 uses a topic “similarity“ Rd (s1, s2) ”between the utterance set template“ template 1 ”and the utterance set template“ template 6 ”with the same topic“ “1” because it is “food”. Similarly, the similarity calculation unit 512 changes the similarity “Rd (s1, s2)” according to the topic between the utterance set template “template 1” and the utterance set template “template 5” to different “food”, “ “Human” is set to “0”. The reason for calculating the similarity based on the topic is to prevent the conversations from being similar to each other by continuing the utterances of the same topic.

  The case where the similarity is calculated in the state of the utterance set template has been described so far, but the case where the similarity is calculated in the state of the utterance set will be described below. For example, the similarity calculation unit 512 calculates the similarity between the first utterance and the second utterance between utterance sets. As one aspect of calculating the similarity, an edit distance is taken as an example. The edit distance refers to the number of words representing the number of rewritten words when rewriting from one sentence to another.

In the following, the following two utterance sets are taken as an example.
Utterance set s1
"First utterance: Do you like curry? Second utterance: Curry is hard?"
Utterance set s2
"First utterance: Do you like ice cream? Second utterance: Ice cream is cold"

Specifically, the similarity calculation unit 512 obtains the edit distance between the first utterances included in the utterance set s1 and the utterance set s2.
First utterance of utterance set s1: curry / ha / do you like?
First utterance of utterance set s2: Ice cream / Ha / Do you like it?
In the above example, when rewriting from the first utterance of the utterance set s1 to the first utterance of the utterance set s2, it is only necessary to rewrite “curry” with “ice cream”, so the number of steps becomes “1”. Also, the edit distance often increases as the sentence becomes longer. For this reason, it normalizes with the length of a sentence. However, particles and auxiliary verbs do not affect the content of the sentence, so the normalized edit distance “number of words / sentence length” is obtained by limiting to other words. That is, the similarity calculation unit 512 obtains a normalized editing distance “1/2 = 0.5” for words other than “ha”.

Subsequently, the similarity calculation unit 512 obtains an edit distance between the second utterances included in the utterance set s1 and the utterance set s2.
The second utterance of the utterance set s1: curry / te / spicy / Yone The second utterance of the utterance set s2: Ice cream / ha / cold / Yone In the above example, from the second utterance of the utterance set s1, the utterance set s2 When rewriting to the second utterance, “curry” should be rewritten to “ice cream” and “spicy” should be rewritten to “cold”, so the number of steps becomes “2”. That is, the similarity calculation unit 512 obtains a normalized editing distance “2/2 = 1”.

  Thereafter, the similarity calculation unit 512 obtains the edit distance “0.5 + 1 = 1.5” between the utterance set s1 and the utterance set s2. That is, the edit distance between utterance sets is obtained by the sum of the edit distance between the first utterances and the edit distance between the second utterances. Then, the similarity calculation unit 512 calculates the similarity “Re (s1, s2)” based on the edit distance between the utterance set s1 and the utterance set s2, and the reciprocal of the obtained sum “1 / 1.5 = 0.67. "

  By the way, in the above example, it is determined whether or not the words match, but the relationship between the words may be further considered. An example of the relationship between words is a conceptual difference between words. For the word concept, information stored in the concept dictionary storage unit 205 is used. For example, it is assumed that “curry” is given the concept of three levels of “product”, “food”, and “cooking”. In addition, “ice cream” is given a concept of three levels of “product”, “food”, and “confectionery”. Similarly, “spicy” is given the concept of three levels of “natural”, “natural”, and “taste”. Further, “cold” is given the concept of three levels of “relation”, “quantity”, and “cool”. It can be said that the words that match up to the lower layers of the concept are closer in meaning to each other. Therefore, the edit distance is obtained by the number that does not match the concept.

  Specifically, the similarity calculation unit 512 obtains the edit distance of the concept of the words between the first utterances included in the utterance set s1 and the utterance set s2. In the above example, the concept of the words “curry” and “ice cream” included in each of the first utterances matches up to two levels, so the number of moves is “3-2 = 1”. As a result, the normalized edit distance of the concept of the words between the first utterances is “1/2 = 0.5”.

  Subsequently, the similarity calculation unit 512 obtains the edit distance of the concept of the words between the second utterances included in the utterance set s1 and the utterance set s2. In the above example, since the concepts of the words “curry” and “ice cream” included in each of the second utterances match up to two levels, the number of moves is “3-2 = 1”. In addition, since the concept of the words “spicy” and “cold” included in each of the second utterances does not match even one layer, the number of moves becomes “3-0 = 3”. As a result, the normalized editing distance of the concept of words between the second utterances is “(1 + 3) / 2 = 2”.

  Thereafter, the similarity calculation unit 512 obtains the edit distance “0.5 + 2 = 2.5” between the utterance set s1 and the utterance set s2. Then, the similarity calculation unit 512 calculates the similarity “Rec (s1, s2)” based on the edit distance between the utterance set s1 and the utterance set s2, and the reciprocal of the obtained sum “1 / 2.5 = 0.4. "

  By the way, in the above example, the difference in concept between words is given as an example of the relationship between words, but co-occurrence between words may be further considered. For co-occurrence, the co-occurrence dictionary storage unit 513 is used. The co-occurrence dictionary storage unit 513 stores the co-occurrence rate of any two words, for example. The co-occurrence rate represents how much two words are used in the same document. Such a co-occurrence rate calculation method includes the following (Equation 1) as one aspect.

(Equation 1)

  In (Equation 1), “cosine coefficient” is given as an example of the co-occurrence rate, and “co-occurrence frequency of word A and word B” is the frequency with which word A and word B are used in the same document. means. As described above, the co-occurrence dictionary storage unit 513 stores the co-occurrence rate of two words calculated in advance using data including a large amount of documents.

  Specifically, the similarity calculation unit 512 obtains the edit distance of the word co-occurrence rate between the first utterances included in the utterance set s1 and the utterance set s2. In the above example, the co-occurrence rate of the words “curry” and “ice cream” included in each first utterance is “0.2”. As a result, the normalized editing distance of the word co-occurrence rate between the first utterances is “1 / 0.2 / 2 = 10”.

  Subsequently, the similarity calculation unit 512 obtains the edit distance of the concept of the words between the second utterances included in the utterance set s1 and the utterance set s2. In the above example, the co-occurrence rate of the words “curry” and “ice cream” included in each second utterance is “0.2”. In addition, the co-occurrence rate of the words “spicy” and “cold” included in each second utterance is “0.01”. As a result, the normalized editing distance of the word co-occurrence rate between the second utterances is “(1 / 0.2 + 1 / 0.01) /2=52.5”.

  Thereafter, the similarity calculation unit 512 obtains the edit distance “10 + 52.5 = 62.5” between the utterance set s1 and the utterance set s2. Then, the similarity calculation unit 512 calculates the similarity “Res (s1, s2)” based on the edit distance using the co-occurrence rate between the utterance set s1 and the utterance set s2, and the reciprocal of the obtained sum “1/62”. .5 = 0.016 ".

  Based on the above, the similarity calculation unit 512 calculates the similarity “R (s1, s2)” between the utterance set s1 and the utterance set s2 by the following (Equation 2).

(Equation 2)
R (s1, s2) = Wt * Rt (s1, s2) + Wu * Ru (s1, s2) + Wd * Rd (s1, s2) + We * Re (s1, s2) + Wec * Rec (s1, s2) + Wes * Res (S1, s2)

  “Wt”, “Wu”, “Wd”, “We”, “Wec”, and “Wes” included in (Equation 2) are weights for the respective similarities and take values between 0 and 1. In calculating the similarity “R (s1, s2)”, any of the above similarities may be used.

  The determination unit 514 determines the order of utterance sets or utterance set templates so that the similarity between consecutive utterance sets or between utterance set templates is highest within a range in which the similarity does not exceed a predetermined value. Here, the order determination processing according to the fifth embodiment will be described with reference to FIG. FIG. 14 is a diagram for explaining the order determination process according to the fifth embodiment.

  In FIG. 14, similarities between “utterance set 1”, “utterance set 2”, “utterance set 3”, and “utterance set 4” calculated by the similarity calculation unit 512 are shown. Here, a case where the predetermined value is “0.9” is taken as an example. For example, as illustrated in FIG. 14, the determination unit 514 sets an utterance set to be used next to “utterance set 1” to “utterance set 1” within a range not exceeding a predetermined value “0.9”. The “speech set 4” having the highest similarity is determined.

  Then, the determination unit 514 sets the utterance set used next to the “utterance set 4” to the “utterance” having the highest degree of similarity with the “utterance set 4” within a range not exceeding the predetermined value “0.9”. Set 3 ”. Subsequently, the determination unit 514 sets the utterance set to be used next to the “utterance set 3” to the highest degree of similarity with the “utterance set 3” within a range not exceeding the predetermined value “0.9”. It determines to utterance set 2 ". That is, when the similarity shown in FIG. 14 is the predetermined value “0.9”, the determination unit 514 “utterance set 1”, “utterance set 4”, “utterance set 3”, “utterance set 2” The order of utterance sets is determined in the following order.

  As described above, a predetermined value is set for the degree of similarity between consecutive utterance sets, and utterance sets whose similarity is equal to or higher than the predetermined value are not used continuously. However, if utterance sets with low similarity are used consecutively, the content of the conversation may change suddenly, so that the similarity between consecutive utterance sets is the highest within a range not exceeding the predetermined value. The order of utterance sets is determined.

  The utterance set acquisition unit 502 acquires an utterance set according to the determined order. For example, the utterance set acquisition unit 502 acquires an utterance set from the utterance set storage unit 101 in accordance with the order of the utterance set determined by the determination unit 514. The first output unit 103a outputs the first utterance in the order of acquisition of the utterance set. For example, the first output unit 103a outputs the first utterance included in the utterance set to a predetermined display output device in the order of acquisition of the utterance set by the utterance set acquisition unit 502. The second output unit 103b outputs a second utterance corresponding to the first utterance output by the first output unit 103a to a predetermined display output device after the detection unit 104 detects the user utterance.

  Next, the flow of dialogue processing according to the fifth embodiment will be described with reference to FIG. FIG. 15 is a flowchart illustrating an example of a flow of interactive processing according to the fifth embodiment.

  For example, as shown in FIG. 15, the similarity calculation unit 512 calculates the similarity between utterance sets or between templates of an utterance set (step S501). Further, the determination unit 514 sets the utterance set so that the similarity between consecutive utterance sets or between templates of the utterance set is the highest within a range in which the similarity calculated by the similarity calculation unit 512 does not exceed a predetermined value. Alternatively, the order of the utterance set templates is determined (step S502).

  Further, the utterance set acquisition unit 502 acquires the utterance set stored in the utterance set storage unit 101 according to the order determined by the determination unit 514 (step S503). In addition, the first output unit 103a outputs the first utterance included in the utterance set acquired by the utterance set acquisition unit 502 to a predetermined display output device (step S504).

  Moreover, the detection part 104 determines whether the user's utterance with respect to the 1st utterance output by the 1st output part 103a was detected (step S505). At this time, when the utterance by the user is detected by the detection unit 104 (Yes in step S505), the second output unit 103b corresponds to the first utterance output by the first output unit 103a, and the utterance set acquisition unit 502 The second utterance included in the utterance set acquired in step S506 is output to a predetermined display output device (step S506). After the output of the second utterance, the utterance set acquisition unit 502 acquires the next utterance set from the utterance set storage unit 101 according to the order determined by the determination unit 514 (step S503).

  On the other hand, when the utterance by the user is not detected (No at Step S505), the detection unit 104 determines whether or not a predetermined time has elapsed (Step S507). At this time, when the predetermined time has not elapsed (No at Step S507), the detection unit 104 determines again whether or not the user's utterance has been detected (Step S505). On the other hand, when the predetermined time has elapsed (Yes at Step S507), the utterance set acquisition unit 502 acquires the next utterance set from the utterance set storage unit 101 according to the order determined by the determination unit 514 (Step S503).

  According to this embodiment, since the order of utterance sets is determined based on the similarity between utterance sets, a more natural flow of conversation can be realized.

(Sixth embodiment)
FIG. 16 is a block diagram illustrating a configuration example of the interactive apparatus according to the sixth embodiment. In the sixth embodiment, functional units that perform the same processes as those in the first embodiment, the second embodiment, or the fifth embodiment are denoted by the same reference numerals, and the description of the same processes is described. May be omitted.

  For example, as shown in FIG. 16, the dialogue apparatus 600 includes an utterance set storage unit 101, an utterance set acquisition unit 602, an output unit 103, a detection unit 104, a concept dictionary storage unit 205, and an utterance template storage unit 506. A concept acquisition unit 207, an utterance template acquisition unit 208, an utterance set generation unit 209, a similarity calculation unit 612, a co-occurrence dictionary storage unit 513, and a determination unit 614. Similarly to the first embodiment, the dialogue apparatus 600 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The similarity calculation unit 612 calculates the similarity between utterance sets including a new utterance set or between templates of an utterance set. Since the processing by the similarity calculation unit 612 is the same as the processing by the similarity calculation unit 512 according to the fifth embodiment, detailed description thereof is omitted here.

  The determination unit 614 groups utterance sets having similarities equal to or higher than a predetermined value or templates of utterance sets into the same group, sequentially selects different groups having the closest similarity, and successively sets utterance sets or utterances. Determine the order of the templates in the set. For example, the determination unit 614 groups speech sets based on a rule that the similarity calculated by the similarity calculation unit 612 is equal to or greater than a predetermined value “X1”.

  And the determination part 614 calculates | requires the similarity of groups. The similarity between groups is, for example, one of an average value, a maximum value, and a minimum value of the similarity between utterance sets belonging to the same group. Subsequently, the determination unit 614 does not select a continuous utterance set from the same group, that is, sequentially selects a continuous utterance set from a different group, and determines the order of the utterance sets. However, when the determination unit 614 sequentially selects continuous speech sets from different groups, the determination unit 614 sequentially selects continuous speech sets from the group having the closest similarity between the groups.

  When selecting an utterance set from a group, a group once selected may be selected again only after all groups are selected. Further, it is not necessary to select an utterance set belonging to a group having a similarity less than a predetermined value in each utterance set, and select an utterance set continuously from the same group N times (“N” is a natural number), Then, you may make it select N times from another group similarly.

  The utterance set acquisition unit 602 acquires the utterance set according to the determined order. For example, the utterance set acquisition unit 602 acquires the utterance set from the utterance set storage unit 101 according to the order of the utterance sets determined by the determination unit 614. The first output unit 103a outputs the first utterance in the order of acquisition of the utterance set. For example, the first output unit 103a outputs the first utterance included in the utterance set to a predetermined display output device in the order of acquisition of the utterance set by the utterance set acquisition unit 602. The second output unit 103b outputs a second utterance corresponding to the first utterance output by the first output unit 103a to a predetermined display output device after the detection unit 104 detects the user utterance.

  Next, the flow of dialogue processing according to the sixth embodiment will be described with reference to FIG. FIG. 17 is a flowchart illustrating an example of the flow of interactive processing according to the sixth embodiment.

  For example, as shown in FIG. 17, the similarity calculation unit 612 calculates the similarity between utterance sets or between templates of an utterance set (step S601). The determination unit 614 groups utterance sets whose similarity calculated by the similarity calculation unit 612 is equal to or greater than a predetermined value into the same group (step S602). Subsequently, the determination unit 614 selects a continuous utterance set from each group based on the similarity between the groups, and determines the order of the utterance sets (step S603).

  Further, the utterance set acquisition unit 602 acquires the utterance set stored in the utterance set storage unit 101 according to the order determined by the determination unit 614 (step S604). Also, the first output unit 103a outputs the first utterance included in the utterance set acquired by the utterance set acquisition unit 602 to a predetermined display output device (step S605).

  In addition, the detection unit 104 determines whether or not an utterance by the user with respect to the first utterance output by the first output unit 103a has been detected (step S606). At this time, when the utterance by the user is detected by the detection unit 104 (Yes in step S606), the second output unit 103b corresponds to the first utterance output by the first output unit 103a, and the utterance set acquisition unit 602. The second utterance included in the utterance set obtained by the above is output to a predetermined display output device (step S607). In addition, after outputting the second utterance, the utterance set acquisition unit 602 acquires the next utterance set from the utterance set storage unit 101 in accordance with the order determined by the determination unit 614 (step S604).

  On the other hand, when the utterance by the user is not detected (No at Step S606), the detection unit 104 determines whether a certain time has elapsed (Step S608). At this time, when the predetermined time has not elapsed (No at Step S608), the detection unit 104 determines again whether or not the user's utterance has been detected (Step S606). On the other hand, when the predetermined time has elapsed (Yes at Step S608), the utterance set acquisition unit 602 acquires the next utterance set from the utterance set storage unit 101 according to the order determined by the determination unit 614 (Step S604).

  According to the present embodiment, groups having similarities between utterance sets are grouped, and the order of the utterance sets is determined based on the similarity between the groups, so that more natural dialogue can be realized. it can.

(Seventh embodiment)
FIG. 18 is a block diagram illustrating a configuration example of the interactive apparatus according to the seventh embodiment. In the seventh embodiment, the same reference numerals are given to the functional units that perform the same processes as those in the first embodiment, the second embodiment, the fourth embodiment, or the fifth embodiment, and the same The description of the processing may be omitted.

  For example, as shown in FIG. 18, the dialogue apparatus 700 includes an utterance set storage unit 101, an utterance set acquisition unit 702, an output unit 103, a detection unit 104, a concept dictionary storage unit 205, and an utterance template storage unit 206. A concept acquisition unit 407, an utterance template acquisition unit 208, an utterance set generation unit 409, an extraction unit 411, a similarity calculation unit 712, a co-occurrence dictionary storage unit 513, and a determination unit 714. Similarly to the first embodiment, the dialogue apparatus 700 realizes a dialogue with the user by outputting characters or voices to a predetermined display output device including a display or a speaker.

  The similarity calculation unit 712 calculates the similarity between the extracted words and utterance sets including the new utterance set and between utterance sets. For example, the similarity calculation unit 712 calculates the similarity between the words extracted by the speech recognition processing by the extraction unit 411 and each utterance set, and calculates the similarity between the utterance sets. Since the calculation of the similarity between utterance sets is the same as in the above-described embodiment, the description thereof will be omitted. Here, the calculation of the similarity between the extracted word and the utterance set will be described.

In the following, the following words and utterance sets are given as examples.
The word k1 “curry”
Utterance set s1
"First utterance: Do you like ice cream? Second utterance: Ice cream is cold"

  Specifically, the similarity calculation unit 712 calculates the similarity “Rec (k1, s1) = 1” of the concepts “curry” and “ice cream” included in the word k1 and the utterance set s1. When a plurality of words are included in the utterance set, any one of the average value, the maximum value, and the minimum value may be set as the similarity.

  In addition, the similarity calculation unit 712 has a co-occurrence rate “0.2” of the words “curry” and “ice cream”, a co-occurrence rate “0.01” of the words “curry” and “cold”, and the word “curry”. The co-occurrence rate “0.05” of “curry” and “like” is acquired from the co-occurrence dictionary storage unit 513. Then, the similarity calculation unit 712 calculates the similarity “Res (k1, s1) = (0.2 + 0.01 + 0.05) /3=0.087” using the co-occurrence rate. Here, an example in which the average value is used has been described, but either the maximum value or the minimum value may be used as the similarity.

  Based on the above, the similarity calculation unit 712 calculates the similarity “R (k1, s1)” between the word k1 and the utterance set s1 by the following (Equation 3).

(Equation 3)
R (k1, s1) = Wec * Rec (k1, s1) + Wes * Res (k1, s1)

  “Wec” and “Wes” included in (Equation 3) are weights for the respective similarities, and take values between 0 and 1. In calculating the similarity “R (k1, s1)”, any of the above similarities may be used.

  The determination unit 714 sets the utterance set having the highest similarity as the first utterance set, and the similarity between consecutive utterance sets is highest within a range in which the similarity does not exceed a predetermined value from the first utterance set. Thus, the order of the utterance set is determined. For example, the determination unit 714 sets the utterance set having the highest similarity among the similarities between the word k1 calculated by the similarity calculation unit 712 and each utterance set as the first utterance set.

  Based on the similarity between utterance sets calculated by the similarity calculation unit 712, the determination unit 714 determines the similarity between consecutive utterance sets within a range in which the similarity does not exceed a predetermined value from the first utterance set. The order of the utterance set is determined so that the degree becomes the highest. That is, after the first utterance set is determined, a predetermined value is set for the similarity between consecutive utterance sets, and an utterance set having a similarity greater than or equal to the predetermined value is not used continuously. However, if utterance sets with low similarity are used consecutively, the content of the conversation may change suddenly, so that the similarity between consecutive utterance sets is the highest within a range not exceeding the predetermined value. The order of utterance sets is determined.

  The utterance set acquisition unit 702 acquires the utterance set according to the determined order. For example, the utterance set acquisition unit 702 acquires the utterance set from the utterance set storage unit 101 according to the order of the utterance set determined by the determination unit 714. The first output unit 103a outputs the first utterance in the order of acquisition of the utterance set. For example, the first output unit 103a outputs the first utterance included in the utterance set to a predetermined display output device in the order of acquisition of the utterance set by the utterance set acquisition unit 702. The second output unit 103b outputs a second utterance corresponding to the first utterance output by the first output unit 103a to a predetermined display output device after the detection unit 104 detects the user utterance. If a new word is extracted by the extraction unit 411, the above process is executed again to update the order of the utterance set.

  Next, the flow of interactive processing according to the seventh embodiment will be described with reference to FIG. FIG. 19 is a flowchart illustrating an example of the flow of interactive processing according to the seventh embodiment.

  For example, as shown in FIG. 19, when the utterance by the user is detected by the detection unit 104 (Yes in step S701), the extraction unit 411 executes a speech recognition process, thereby detecting a word included in the utterance by the user. Extract (step S702). Further, when the detection unit 104 has not detected the user's utterance (No in step S701), the extraction unit 411 enters a state of waiting for the user's detection of the utterance by the detection unit 104.

  Further, the similarity calculation unit 712 calculates the similarity between the words extracted by the extraction unit 411 and each utterance set and between utterance sets (step S703). Also, the determination unit 714 determines the first utterance set based on the similarity between the word calculated by the similarity calculation unit 712 and the utterance set, and between the utterance sets calculated by the similarity calculation unit 712. The order of the utterance set following the first utterance set is determined based on the similarity (step S704).

  In addition, the utterance set acquisition unit 702 acquires the utterance set stored in the utterance set storage unit 101 in the order determined by the determination unit 714 (step S705). Also, the first output unit 103a outputs the first utterance included in the utterance set acquired by the utterance set acquisition unit 702 to a predetermined display output device (step S706).

  Moreover, the detection part 104 determines whether the user's utterance with respect to the 1st utterance output by the 1st output part 103a was detected (step S707). At this time, when the user's utterance is detected by the detection unit 104 (Yes in step S707), the second output unit 103b corresponds to the first utterance output by the first output unit 103a, and the utterance set acquisition unit 702 The second utterance included in the utterance set acquired in step S708 is output to a predetermined display output device (step S708). After the output of the second utterance, the utterance set acquisition unit 702 acquires the next utterance set from the utterance set storage unit 101 according to the order determined by the determination unit 714 (step S705).

  On the other hand, when the utterance by the user is not detected (No at Step S707), the detection unit 104 determines whether or not a certain time has elapsed (Step S709). At this time, when the predetermined time has not elapsed (No at Step S709), the detection unit 104 determines again whether or not the user's utterance has been detected (Step S707). On the other hand, when the predetermined time has elapsed (Yes in step S709), the utterance set acquisition unit 702 acquires the next utterance set from the utterance set storage unit 101 according to the order determined by the determination unit 714 (step S705).

  According to the present embodiment, since an utterance set including a word extracted by speech recognition is used, a more natural dialogue can be realized.

(Embodiments other than the above)
In the above-described embodiment, the case where the dialogue is realized without recognizing the content of the utterance by the user has been described. However, it is preferable to realize the dialogue that does not hinder the continuation of the dialogue. For example, when the utterance by the user is a predetermined utterance such as “return”, if the next utterance is output as it is, there is a possibility that the dialogue is ignored by the user. In order to avoid this, when the utterance by the user is a predetermined utterance such as “return”, the previous utterance can be output again.

  FIG. 20 is a flowchart illustrating an example of the flow of dialogue processing related to a request for recurrent speech. For example, as shown in FIG. 20, the interactive device 100 acquires an utterance set from the utterance set storage unit 101 (step S801), and outputs the first utterance included in the acquired utterance set to a predetermined display output device ( Step S802).

  Then, when the utterance by the user is detected (Yes at Step S803), the dialogue apparatus 100 determines whether or not the detected utterance by the user is a request for re-utterance (Step S804). In determining whether or not it is a request for such a recurrent utterance, a predetermined utterance such as “speak again” is held in advance, and it is determined whether or not the detected utterance by the user corresponds to the predetermined utterance. Further, when a predetermined button is pressed, this may be regarded as a request for re-speech. At this time, when the interactive apparatus 100 determines that the request is for a recurrent utterance (Yes at step 804), it outputs the first utterance again (step S802). In the re-output of the first utterance, for example, the voice may be increased or the utterance speed may be further reduced so as to be easily understood by the user.

  On the other hand, when determining that the request is not a re-utterance request (No at Step S804), the interactive apparatus 100 outputs the second utterance (Step S805). In addition, when the utterance by the user is not detected (No at Step S803), the dialogue apparatus 100 determines whether a certain time has passed (Step S806). At this time, when it is determined that the predetermined time has not elapsed (No at Step S806), the interactive apparatus 100 executes the process at Step S803. On the other hand, when it is determined that the certain time has elapsed (Yes at Step S806), the interactive apparatus 100 executes the process at Step S801. The re-utterance request determination process may be executed even after the second utterance is output.

  According to the present embodiment, when the user utterance is a predetermined utterance, since the immediately previous utterance is output again, it is possible to suppress a one-way conversation.

  Further, in the above embodiment, the case has been described in which the utterance using the next utterance set is performed if the utterance by the user cannot be detected for a certain period of time, but an appropriate utterance may be made even if the user does not utter anything. it can. Here, a case where the “second utterance” representing the utterance of the response of the utterance by the user includes the “third utterance” representing the utterance output when the user does not utter anything.

  FIG. 21 is a diagram illustrating an example of information stored in the utterance set storage unit 101 including the third utterance. For example, as illustrated in FIG. 21, the utterance set storage unit 101 includes a first utterance “Is a movie?”, An assumed user utterance “I like you”, “(no response)”, and a second utterance “movie is I remember "Speech Set 1" including "I like movies" (3rd utterance). Similar to the above-described embodiment, the “assumed user utterance” may not be included in the utterance set storage unit 101.

  FIG. 22 is a flowchart illustrating an example of the flow of interactive processing when a user utterance cannot be detected at a certain time. For example, as shown in FIG. 22, the dialogue apparatus 100 acquires an utterance set from the utterance set storage unit 101 (step S901), and outputs the first utterance included in the acquired utterance set to a predetermined display output device ( Step S902).

  Then, when the utterance by the user is detected (Yes at Step S903), the dialogue apparatus 100 determines whether or not the detected utterance by the user is a request for re-utterance (Step S904). At this time, when the dialogue apparatus 100 determines that the request is a request for recurrent utterance (Yes in step S904), it outputs the first utterance again (step S902). On the other hand, when determining that the request is not a re-utterance request (No at step S904), the interactive apparatus 100 outputs the second utterance (step S905).

  In addition, when the utterance by the user is not detected (No at Step S903), the dialogue apparatus 100 determines whether a certain time has passed (Step S906). At this time, when it is determined that the predetermined time has not elapsed (No at Step S906), the interactive apparatus 100 executes the process at Step S903. On the other hand, when determining that the predetermined time has elapsed (Yes at Step S906), the interactive apparatus 100 outputs the third utterance (Step S907).

  According to the present embodiment, even when a user utterance cannot be detected at a certain time, a corresponding utterance is output, so that an appropriate dialogue can be realized.

  In the second embodiment, the concept dictionary storage unit 205 does not include proper nouns or new words, but unknown words can be added in accordance with the concepts described in the utterance set template. . When an unknown word is added, the concept of the word is acquired and assigned from the concept dictionary storage unit 205, the utterance set template including the concept is searched, and the utterance set is generated. This makes it possible to add not only commonly used words but also hot keywords and recommended product names. For example, an unknown word with the product name “Chocolate Crunch NEW” is set as a concept “food”, and a phrase “delicious enough to become addictive” that is a feature of the product is set as a concept “taste”. As a result, it is possible to generate an utterance set such as “Do you like chocolate crunch NEW?” Or “Chocolate crunch NEW is delicious enough to be addictive”. That is, according to the present embodiment, the present invention can be applied not only to daily conversation but also to sales promotion of products.

  Moreover, although the said embodiment demonstrated the case where similarity was calculated based on the edit distance of the 1st utterances contained in two utterance sets, and the 2nd utterances, a 2nd utterance and the following utterance set of The similarity can also be calculated based on the edit distance from the first utterance. That is, since the user recognizes in the order of the first utterance and the second utterance, it is not preferable that the utterance between the second utterance and the first utterance of the next utterance set is similar. The similarity is calculated based on the above.

  Further, the above-described embodiments are presented as examples, and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 100 Dialogue device 101 Utterance set memory | storage part 102 Utterance set acquisition part 103 Output part 103a 1st output part 103b 2nd output part 104 Detection part

Claims (9)

An utterance set storage unit that stores an utterance set including a first utterance and a second utterance representing an utterance of a response to an utterance by a user assumed as a response to the first utterance;
An utterance set acquisition unit for acquiring the utterance set;
A first output unit that outputs the first utterance included in the acquired utterance set;
A detection unit for detecting an utterance by the user after the first utterance is output;
And a second output unit that outputs the second utterance included in the acquired utterance set when an utterance by the user is detected. An utterance template storage unit for storing a template of the utterance set in which some words are expressed in concept;
A concept dictionary storage that stores the word and the concept in association with each other;
A concept acquisition unit that acquires the concept corresponding to the word from the concept dictionary storage unit when the word is input;
An utterance template acquisition unit for acquiring a template of the utterance set including the acquired concept from the utterance template storage unit;
An utterance set generation unit that generates a new utterance set by inserting the input phrase into the concept included in the acquired utterance set template, and stores the generated new utterance set in the utterance set storage unit The interactive apparatus according to claim 1, further comprising: An extractor for extracting a phrase included in the utterance by the user that matches the phrase stored in the concept dictionary storage;
The concept acquisition unit acquires the concept corresponding to the extracted phrase from the concept dictionary storage unit,
The dialogue apparatus according to claim 2, wherein the utterance set generation unit generates a new utterance set by inserting the extracted phrase into a concept included in the acquired template of the utterance set. . An utterance template storage unit for storing a template of the utterance set in which some words are expressed in concept;
A unique word storage unit that stores a unique phrase acquired from the outside and the concept of the unique phrase in association with each other;
An utterance template acquisition unit for acquiring a template of the utterance set including the concept of the unique phrase from the utterance template storage unit;
An utterance set generation unit that generates a new utterance set by inserting the unique words into the concept included in the acquired template of the utterance set, and stores the generated new utterance set in the utterance set storage unit; The interactive apparatus according to claim 1, further comprising: A similarity calculation unit that calculates a similarity between the utterance sets including the new utterance set, or between templates of the utterance set;
Determination that determines the order of the utterance set or the templates of the utterance set so that the similarity between the consecutive utterance sets or between templates of the utterance set is the highest within a range where the similarity does not exceed a predetermined value. And further comprising
The utterance set acquisition unit acquires the utterance set according to the order,
The interactive apparatus according to claim 2, wherein the first output unit outputs the first utterance in the order of acquisition of the utterance set. A similarity calculation unit that calculates a similarity between the utterance sets including the new utterance set, or between templates of the utterance set;
Grouping the utterance sets or the utterance set templates whose similarity is equal to or greater than a predetermined value into the same group, sequentially selecting different groups with the closest similarity, and the successive utterance sets or utterance sets And a determination unit for determining the order of the templates,
The utterance set acquisition unit acquires the utterance set according to the order,
The interactive apparatus according to claim 2, wherein the first output unit outputs the first utterance in the order of acquisition of the utterance set. A similarity calculation unit that calculates a similarity between the extracted phrase and the utterance set including the new utterance set, and between the utterance sets;
The utterance set having the highest similarity is set as the first utterance set, and the similarity between consecutive utterance sets is highest within a range in which the similarity does not exceed a predetermined value from the first utterance set. And a determination unit for determining the order of the utterance set,
The utterance set acquisition unit acquires the utterance set according to the order,
The interactive apparatus according to claim 3, wherein the first output unit outputs the first utterance in order of acquisition of the utterance set. The detection unit further detects a predetermined utterance by the user,
The interactive apparatus according to claim 1, wherein the first output unit outputs the first utterance when a predetermined utterance by the user is detected. The utterance set storage unit stores the utterance set further including a third utterance representing an utterance of a response to a case where the user does not utter,
9. The dialogue apparatus according to claim 1, wherein the second output unit outputs the third utterance included in the acquired utterance set when an utterance by the user is not detected. .

Images