JP2019220022A - Interactive device, interactive device control method, and program - Google Patents

Abstract

To allow a user to feel as if an utterance content had been growing without increasing the number of pieces of response sentence data.SOLUTION: An interactive device 100 comprises a syntactic analysis unit 11, a model response sentence acquisition unit 12, a word use propriety determination unit 13, a response sentence creation unit 14 and a response unit 15. The syntactic analysis unit 11 subjects a user utterance sentence input from a user to syntactic analysis. The model response sentence acquisition unit 12 acquires a model response sentence which is an exemplary response sentence according to the result of syntactic analysis from the syntactic analysis unit 11. The word use propriety determination unit 13 determines whether each word included in the model response sentence acquired by the model response sentence acquisition unit 12 is usable or unusable according to a predetermined condition. The response sentence creation unit 14 creates a response sentence to the user utterance sentence by removing a word determined as unusable by the word use propriety determination unit 13 from the model response sentence acquired by the model response sentence acquisition unit 12. The response unit 15 responds to the user by using the response sentence created by the response sentence creation unit 14.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an interactive device, an interactive device control method, and a program.

  Interactive devices that interact with the user in response to the user's utterance have been developed. For example, Patent Literature 1 discloses a legged mobile robot that grows an interactive function according to the degree of growth of an internal state.

JP-A-2002-321177

  The legged mobile robot disclosed in Patent Literature 1 stores a plurality of pieces of response sentence data having different intellectual levels from each other, and from the plurality of pieces of response sentence data, generates response sentence data corresponding to the degree of growth of the robot. A response is made based on the selected response sentence data. This allows the interactive function of the robot to grow in accordance with the degree of growth.

  However, as described above, since this robot must store a plurality of response sentence data corresponding to the degree of growth, the number of response sentence data becomes enormous as a whole, and the response sentence data is stored. Storage capacity increases.

  SUMMARY An advantage of some aspects of the invention is to provide a dialogue apparatus and a dialogue apparatus control method that can make a user feel as if the utterance content is growing without increasing the number of response sentence data. And to provide programs.

In order to achieve the above object, the interactive device of the present invention includes:
Parsing means for parsing a user utterance sentence which is an utterance sentence input by the user;
An exemplary response sentence obtaining unit that obtains an exemplary response sentence that is an exemplary response sentence according to a syntax analysis result of the user utterance sentence by the syntax analysis unit;
According to a predetermined condition, word use availability determination means for determining whether or not each word included in the model response sentence acquired by the model response sentence acquisition means is usable or unusable,
Response sentence generation means for generating a response sentence to the user utterance sentence by deleting a word determined to be unusable by the word use availability determination means from the model response sentence acquired by the model response sentence acquisition means,
Response means for responding to the user using the response sentence generated by the response sentence generation means,
Is provided.

  According to the present invention, it is possible to make the user feel that the utterance content is growing without increasing the number of response sentence data.

FIG. 2 is a diagram illustrating an example of an appearance of the interactive device according to the first embodiment of the present invention. FIG. 7 is a diagram illustrating another example of the appearance of the interactive device according to the first embodiment. FIG. 2 is a diagram illustrating a functional configuration of the interactive device according to the first embodiment. FIG. 4 is a diagram illustrating an example of data stored in a word DB of the interactive device according to the first embodiment. FIG. 4 is a diagram illustrating an example of data stored in a model response sentence DB of the interactive device according to the first embodiment. 6 is a flowchart of a dialog process of the dialog device according to the first embodiment. 6 is a flowchart of a word DB update process of the interactive device according to the first embodiment. 9 is a flowchart of a response sentence generation process of the interactive device according to the first embodiment. 5 is a flowchart of a multi-word response sentence generation process of the interactive device according to the first embodiment. FIG. 9 is a diagram illustrating a process of step S401 of a multiple-word response sentence generation process of the interactive device according to the first embodiment. FIG. 8 is a diagram illustrating a process of correcting known information based on dependency information in a multiple-word response sentence generation process of the interactive device according to the first embodiment. FIG. 4 is a diagram illustrating an example of response sentence generation by the interactive device according to the first embodiment. It is a figure showing an example of data memorized by word DB of a dialog device concerning Embodiment 2 of the present invention. FIG. 14 is a diagram illustrating an example of response sentence generation by the interactive device according to the second embodiment. FIG. 14 is a diagram illustrating another example of generating a response sentence by the interactive device according to the second embodiment. It is a figure showing an example of data memorized by word DB of a dialog device concerning Embodiment 3 of the present invention. 13 is a flowchart of a word DB update process of the interactive device according to the third embodiment.

  Hereinafter, a dialog device according to an embodiment of the present invention will be described with reference to the drawings. In the drawings, the same or corresponding parts have the same reference characters allotted.

(Embodiment 1)
In the dialogue device according to the embodiment of the present invention, as the dialogue with the user is repeated, usable words (nouns, verbs, adjectives, etc.) increase, and the dialogue device grows as if by the dialogue with the user. An interactive device that appears to be

  As illustrated in FIG. 1, the interactive device 100 according to the first embodiment has an appearance imitating a human child. A microphone as a voice input unit 31 is provided at an ear position, and a voice output unit 32 is provided at a mouth position. Speaker. FIG. 1 is merely an example of the appearance of the interactive device 100. The interaction device 100 may have a smart speaker-like appearance, for example, as shown in FIG. In FIG. 2, the dialogue device 100 also includes a microphone as the voice input unit 31 and a speaker as the voice output unit 32 in appearance.

  As illustrated in FIG. 3, the interactive device 100 includes a control unit 10, a storage unit 20, an audio input unit 31, an audio output unit 32, a driving unit 33, an operation input unit 34, and a communication unit 35 as functional components.

  The control unit 10 is configured by a CPU (Central Processing Unit) or the like, and executes a program stored in the storage unit 20 to execute each unit (syntax analysis unit 11, model response sentence acquisition unit 12, word availability determination The functions of the unit 13, the response sentence generation unit 14, and the response unit 15) are realized.

  The storage unit 20 is configured by a ROM (Read Only Memory), a RAM (Random Access Memory), or the like, and a part or all of the ROM is configured by an electrically rewritable memory (a flash memory or the like). The storage unit 20 functionally includes a word DB (DataBase) 21 and a model response sentence DB 22. The ROM stores a program to be executed by the CPU of the control unit 10 and data necessary for executing the program in advance. The RAM stores data created or changed during execution of the program.

  As shown in FIG. 4, the words included in the response sentence of the interactive device 100 are stored in the word DB 21 together with the known information of the words. The known information is information indicating whether the sentence input by the user to the interactive device 100 includes the same word as the word of the response sentence. This is for treating a word included in a sentence that the user has at least once input to the interactive device 100 as a word known to the interactive device 100. The known information is usually initialized with an initial value “NO” for all words. The words stored in the word DB 21 are basically nouns, verbs, adjectives, and adjective verbs. In the word DB 21, in principle, the original form (the form described in a general dictionary (the form before utilization)) is referred to as the “original form” for the words to be used. ) Is stored, and a word included in a sentence that the user has once input to the interactive device 100 is treated as a known word even if the word has a shape other than the original form. In addition, nouns, verbs, adjectives, and words other than adjectives (pronouns, adverbs, adjuvants, conjunctions, inflections, adjuncts (particles, auxiliary verbs)) are used at least once in a sentence input by the user to the interactive device 100. Even if it is not included, it will be treated as a known word.

  As shown in FIG. 5, the model response sentence DB 22 stores a model (typical) response sentence to the sentence according to a keyword included in the sentence input to the interactive device 100 by the user.

  The voice input unit 31 includes a microphone, and acquires data (voice data) of voice uttered by the user to the interactive device 100. The control unit 10 executes the voice recognition program stored in the storage unit 20 to perform voice recognition on the voice data acquired by the voice input unit 31 by a known method, and to output a sentence uttered by the user (user uttered sentence). Can be obtained. The voice input unit 31 functions as a voice input unit together with the voice recognition program.

  The voice output unit 32 includes a voice synthesizing unit and a speaker. The voice output unit 32 can synthesize a sentence given from the control unit 10 by using a well-known technique and output (speech) as voice. The audio output unit 32 functions as an audio output unit.

  The drive unit 33 includes a motor and the like, and can tilt the neck and change the facial expression. With the driving unit 33, the interactive device 100 can indicate that the sentence input by the user cannot be understood by an operation such as tilting the head. The driving unit 33 functions as a driving unit.

  The operation input unit 34 includes a switch, a touch panel, and the like, and acquires operation content data (such as a control command for the interactive device 100) input by the user.

  The communication unit 35 is a wireless module including an antenna for wirelessly communicating communication data with an external network or the like. For example, the communication unit 35 is a wireless module for performing wireless communication with another computer via a wireless LAN (Local Area Network). By using the communication unit 35, the interactive device 100 can exchange various information with an external server or the like.

  Next, a functional configuration of the control unit 10 of the interactive device 100 will be described. The control unit 10 realizes the functions of the syntax analysis unit 11, the model response sentence acquisition unit 12, the word use determination unit 13, the response sentence generation unit 14, and the response unit 15, and responds to the utterance of the user.

  The syntax analysis unit 11 performs a syntax analysis of a user utterance sentence, which is a user utterance sentence obtained by the control unit 10 performing voice recognition on the voice data obtained by the voice input unit 31 and using a well-known technique. By this syntax analysis, the control unit 10 can acquire information such as words, phrases, and dependencies included in the user utterance sentence. Further, the syntax analysis unit 11 may parse the model response sentence acquired by the model response sentence acquisition unit 12 to acquire information such as words, phrases, and dependencies included in the model response sentence. it can. The syntax analyzer 11 functions as a syntax analyzer.

  The model response sentence acquisition unit 12 acquires a model response sentence that is a model response sentence to the user utterance sentence that has been parsed by the syntax analysis unit 11. The model response sentence DB22 stores a model response sentence for the user utterance sentence according to the keyword included in the user utterance sentence. An exemplary response sentence can be obtained. Specifically, the model response sentence acquisition unit 12 includes, among words included in the keyword item of the model response sentence DB 22, words included in the user utterance sentence parsed by the syntax analysis unit 11. Whether or not the same word exists is searched, and if it exists, an example response sentence having the word as a keyword is obtained from the example response sentence DB 22. The model response sentence acquisition unit 12 functions as a model response sentence acquisition unit.

  The word availability determination unit 13 determines whether each word included in the model response sentence acquired by the model response sentence acquisition unit 12 is usable or unusable in the response sentence to the user. This determination is made in a response sentence generation process described later, but simply, the word availability determination unit 13 determines that the word whose known information is “YES” in the word DB 21 is available, and the known information is determined. The word "NO" is determined to be unusable. As described above, the item of the known information in the word DB 21 is “YES” for a word included at least once in a past user utterance sentence. Therefore, the word availability determination unit 13 determines that a word that has never been recognized by the interactive device 100 is unusable, and determines that a word that has been recognized at least once is usable. The word availability determination unit 13 functions as a word availability determination unit.

  The response sentence generation unit 14 generates a response sentence to the user utterance sentence by deleting the words determined to be unusable by the word use availability determination unit 13 from the model response sentence acquired by the model response sentence acquisition unit 12. Details of the processing for generating the response sentence will be described later. The response sentence generation unit 14 functions as a response sentence generation unit.

  The response unit 15 responds to the user by outputting the response sentence generated by the response sentence generation unit 14 by voice from the audio output unit 32 using a known method. The response unit 15 functions as a response unit.

  The functional configuration of the interactive device 100 has been described above. The interactive device 100 is placed in an environment (room or the like) where the user lives, and analyzes the utterance sentence uttered by the user and registers a word used in the utterance sentence as a known word. At the same time, it responds to the utterance sentence of the user with a response sentence composed of known words. The process for this is the interactive process described below. The interactive process may be performed at all times while the power of the interactive device 100 is turned on, or performed when a user's instruction is given by an input from the operation input unit 34 or the like. May be.

  The interactive process performed by the interactive device 100 will be described with reference to FIG. First, the control unit 10 determines whether or not to end the interactive processing (Step S101). For example, when the power of the interactive device 100 is turned off or when the end of the interactive process is instructed from the operation input unit 34, the control unit 10 determines that the interactive process is to be ended. If it is determined that the interactive processing is to be ended (Step S101; Yes), the processing is ended.

  If it is determined that the dialog processing is not to be ended (Step S101; No), the control unit 10 acquires a sentence uttered by the user (user uttered sentence) via the voice input unit 31 (Step S102). Then, the syntax analysis unit 11 analyzes the syntax of the user utterance sentence, and acquires words and the like included in the user utterance sentence (step S103). Step S103 is also called a syntax analysis step.

  Next, the control unit 10 performs a word DB update process (step S104). The word DB updating process is a process of updating the contents (especially known information) of the word DB according to the words included in the user utterance sentence, and details of the process will be described later.

  Next, the model response sentence acquisition unit 12 acquires a model response sentence for the user utterance sentence from the model response sentence DB 22 based on the words included in the user utterance sentence (step S105). Step S105 is also called an exemplary response sentence acquisition step. Then, the response sentence generation unit 14 performs a response sentence generation process to generate a response sentence (Step S106). In the response sentence generation process, the words determined to be unusable by the word useability determination unit 13 are deleted from the model response sentence acquired by the model response sentence acquisition unit 12, and the response sentence generation unit 14 responds to the user utterance sentence. The processing for generating a sentence will be described later in detail.

  Then, the response unit 15 utters the response sentence generated by the response sentence generation unit 14 through the voice output unit 32 (step S107). Step S107 is also called a response step. Then, the process returns to step S101.

  The interactive processing has been described above. Next, the word DB updating process performed in step S104 of the interactive process (FIG. 6) will be described with reference to FIG.

  First, the control unit 10 searches the word DB 21 for the same word as the word included in the user utterance sentence acquired in step S102 (step S201). Since the user utterance sentence has been subjected to the syntax analysis in step S103 of the dialog processing (FIG. 6), the control unit 10 can grasp all the words included in the user utterance sentence. In this case, the number of words retrieved from the word DB 21 may be singular or plural depending on the content of the user utterance sentence.

  Next, the control unit 10 reads one word that is the same as the word searched in step S201 from the word DB 21 to check the known information of each word stored in the word DB 21 (step S202). Then, the control unit 10 determines whether or not the read known information of the word is “YES” (Step S203). If the known information is “YES” (Step S203; Yes), the process proceeds to Step S205.

  If the known information is “NO” (Step S203; No), the known information of the word in the word DB 21 is changed to “YES” and the word DB 21 is updated (Step S204). Then, the control unit 10 determines whether all the words searched from the word DB 21 have been confirmed (Step S205). If all are confirmed (step S205; Yes), the word DB update processing is terminated, and the flow proceeds to step S105 of the interactive processing (FIG. 6). If all have not been confirmed yet (step S205; No), the process returns to step S202.

  The word DB update processing has been described above. As a result of the word DB update process, among the words stored in the word DB 21, the known information of the same word as the word included in the user utterance becomes “YES”. As a result, the interactive device 100 comes to recognize the word as a “word you know”. Next, the response sentence generation processing performed in step S106 of the interactive processing (FIG. 6) will be described with reference to FIG.

  First, the syntax analysis unit 11 analyzes the syntax of the model response sentence, and acquires words and the like included in the model response sentence (step S301). Then, the word use determination unit 13 searches the word DB 21 for the same word as the word included in the model response sentence (step S302), and among the searched words, the word whose known information is “YES” It is determined whether there is one or more (step S303). If there is no word whose known information is “YES” among the searched words (Step S303; No), the response sentence generation unit 14 generates an empty sentence containing no word as a response sentence (Step S303). Step S304).

  Then, the control unit 10 causes the driving unit 33 to perform a tilting operation or the like, so that the interactive device 100 responds to the user with a gesture (step S305), ends the response sentence generation process, and performs the interactive process (FIG. 6). The process proceeds to step S107). The response in the gesture in step S305 is not limited to the operation of tilting the neck, and the driving unit 33 responds with various operations indicating that the words cannot be understood, such as giving a troubled expression or a smiling expression. be able to.

  On the other hand, if there is one or more words whose known information is “YES” among the words searched from the word DB 21 (Step S303; Yes), the word use determination unit 13 determines that only one of the searched words is used. It is determined whether or not (step S306). If the searched word is not only one (step S306; No), the response sentence generation unit 14 performs a multiple word response sentence generation process to generate a response sentence (step S308), and executes the response sentence generation process. After ending, the process proceeds to step S107 of the interactive process (FIG. 6). The multi-word response sentence generation process is a process of generating a response sentence including a plurality of words, and the details of the process will be described later.

  On the other hand, if there is one word whose known information is “YES” among the words searched from the word DB 21 (step S306; Yes), the response sentence generation unit 14 generates a response sentence including only the same word as the word. It is generated (step S307).

  The response sentence generation processing has been described above. Next, the multi-word response sentence generation process performed in step S308 of the response sentence generation process (FIG. 8) will be described with reference to FIG.

  First, the word-usability determining unit 13 divides the model response sentence after the syntax analysis into segments, and initializes known information of each segment (step S401). For example, as shown in FIG. 10, when the model response sentence is “Autumn fish is delicious if it is greasy with the taste of autumn, it is delicious.” / It's tasty / it's delicious. " Then, the known information of the word included in each phrase is obtained by referring to the word DB 21, and the known information of the phrase is initialized as the known information of the phrase. For example, the phrase "Swordfish" includes the word "Swordfish", and the known information is "YES" in the example of FIG. 10, so the known information of the phrase "Swordfish" is "YES". Initialized as

  Also, the phrase “autumn” includes the word “autumn”, but since the known information is “NO” in the example of FIG. 10, the known information of the phrase “autumn” is “ "NO". In this way, known information is set for all the phrases of the model response sentence. When one phrase includes a plurality of words, if the known information of all the words is “YES”, the known information of the phrase is initialized to “YES”, and even one of the known information is Contains the word "NO", the known information of the phrase is initialized to "NO".

  Next, based on the dependency information of each phrase of the model response sentence, the word useability determination unit 13 determines the known information of each phrase and, if the known information of the phrase to which the phrase is modified is “NO”, , "NO". The correction of the known information is performed in the processing of steps S402 to S406 of the multiple word response sentence generation processing (FIG. 9). It should be noted that the modification may be performed in two or more stages. In the following processing, the modification information is a processing for checking the modification source of each clause from the end of the sentence to the beginning of the sentence.

  First, the word-usability judging unit 13 extracts the phrases one by one from the end of the model response sentence toward the beginning of the sentence in order to confirm the modification source of each phrase (step S402). Then, the word use determination section 13 determines whether or not the extracted phrase has a modification source (step S403). If the phrase has no modifier (step S403; No), the process proceeds to step S406. If the phrase has a modification destination (step S403; Yes), it is determined whether the known information of the phrase extracted this time is “YES” (step S404). If the known information of the phrase extracted this time is “YES” (Step S404; Yes), the process proceeds to Step S406.

  If the known information of the phrase extracted this time is not “YES” (step S404; No), the word useability determination unit 13 sets the known information of the modification source phrase of the phrase extracted this time to “NO” (step S405). ), And proceed to step S406. For example, as shown in FIG. 11, the phrase at the end of the model response sentence “delicious,” which is the modifier, is “Notori,” but the known information of “delicious” is “YES,” so nothing is performed. The process proceeds to step S406. The original phrase of the next clause "Noto" is "Fataga", but the known information of "Noto" is "NO", so the known information of "Fataga" is also set to "NO". You.

  Returning to FIG. 9, in step S406, the word use determination unit 13 determines whether or not the modification destinations of all the phrases included in the model response sentence have been confirmed. If all the clauses have not been confirmed (Step S406; No), the process returns to Step S402. If all the clauses have been confirmed (step S406; Yes), the response sentence generation unit 14 generates a response sentence by arranging the clauses whose known information is “YES” (step S407). For example, in the example shown in FIG. 11, finally, the known information is “YES” in only two phrases, “the swordfish” and “delicious”. By arranging the two phrases, a response sentence saying "Autumn fish is delicious" is generated.

  The multi-word response sentence generation processing has been described above. By the response sentence generation processing and the multi-word response sentence generation processing described above, the response sentence generation unit 14 is configured from the model response sentence by a word corresponding to the word whose known information is “YES” at that time. Generate a response sentence. In the response sentence generation process and the multi-word response sentence generation process, the step performed by the word use availability determination unit 13 is also called a word use availability determination step, and the step performed by the response sentence generation unit 14 is also called a response sentence generation step.

  Further, in the above-described multi-word response sentence generation processing, processing is performed on the assumption that all of the dependencies are related to a clause following a sentence in front of a sentence (a modification source is located in front of a sentence and a modification destination is located in the back). , But is not limited to this. Even in the case where there is a dependency in the reverse direction, a process of setting the known information of the modification source phrase to “NO” may be performed when the known information of the modification destination phrase is “NO”. Dependency information as to which clause is the modification destination or modification source of which clause can be obtained from the result of the syntax analysis performed by the syntax analysis unit 11. By performing the above-described processing based on this information, the dependency information can be obtained. Regardless of which direction the receiving direction is, the word use determination unit 13 can correct the known information of the phrase correctly.

  Next, a specific operation example of the interactive device 100 will be described with reference to FIG. First, at the first time, it is assumed that the known information of all the words has been initialized to “NO” in the word DB 21 as shown in FIG. In addition, it is assumed that the model response sentence shown in FIG. 5 is stored in the model response sentence DB 22 together with the keyword. Then, suppose that the user uttered, "I cooked saury at school today, but it was delicious."

  Then, the control unit 10 acquires, as the user's utterance sentence, “I cooked the sword fish at school today, but it was delicious.” The word DB updating process updates the word DB 21 as shown in FIG. 12. In addition, since the keyword “swordfish” is extracted from the words in the user utterance sentence, the model response sentence acquisition unit 12 reads from the model response sentence DB 22 that “swordfish is delicious if it is greasy with the taste of autumn.” Get response statement.

  Then, the response sentence "Autumn fish is delicious" is generated by the above-described response sentence generation processing and the multi-word response sentence generation processing. At this point, the interactive device 100 has not yet recognized words such as “autumn” and “taste”, and therefore responds to the user with a response sentence that does not use these unknown words. Thereafter, when a user utterance sentence including words such as "autumn" and "taste" is obtained, the known information of these words becomes "YES", and thereafter, a response sentence using "autumn" and "taste" is also used. (For example, “Autumn fish is delicious in autumn taste.”), And the user can respond with such a longer response sentence.

  As described above, the dialog device 100 can respond with a more complicated expression as various dialogs with the user are performed without expanding the model response sentence DB 22. You can feel like you are growing.

  Then, the interactive device 100 responds with a response sentence composed of words included in the user utterance sentence, so that the interaction with the interactive device 100 for the user can be as if the child were taught a word. . Therefore, the user can have a special feeling of “I am growing up” or “I grew up in my own home” with respect to the interactive device 100, and the user becomes more attached to the interactive device 100.

  Also, the more the user talks to the interactive device 100, the more he / she can realize that the interactive device 100 grows, and therefore, the user will want to speak various things. For this reason, the interactive device 100 can collect various various information from the user.

  Since the interactive device 100 can generate a meaningful response sentence using a word included in the user utterance sentence from one model response sentence, a large amount of information for individually responding to each growth stage of the interactive device 100 is provided. There is no need to prepare a response statement. If a response sentence for individually responding to each growth stage is prepared, a situation may occur in which the content of the response sentence differs for each growth stage. Then, when such a situation occurs, the user may be given an impression such as “what is said is different from before” or “character has changed”. However, in the interactive device 100, such a situation can be avoided.

(Embodiment 2)
In the dialogue device 100 according to the first embodiment, of the words registered in the word DB 21, the words to be used are registered in principle in their original forms (forms described in a general dictionary; end forms in Japanese grammar). ing. Then, in the interactive apparatus 100, even if a form other than the original form appears in the user utterance sentence, the word is treated as a known word. For this reason, verbs, adjectives and adjective verbs (hereinafter, verbs, adjectives and adjective verbs are collectively referred to as declinable words) that frequently appear in user utterances are immediately known words and the user feels growth. There is a possibility that there is not enough time to be taken. Therefore, words to be utilized are not treated collectively in the original form, but are treated separately according to the utilized form. ) And other forms are treated as different words.) Embodiment 2 will be described.

  The functional configuration of the interactive device 101 according to the second embodiment is the same as the interactive device 100 according to the first embodiment except for the word DB 21. As shown in FIG. 13, the word DB 21 of the interactive device 101 stores the past tense of a verb as a word different from the original. For example, in FIG. 13, the past tense “baked” is stored in the word DB 21 in addition to “baked”, and the past tense “delicious” is stored in addition to “delicious”. In the case of Japanese, the past tense is represented as “a word group consisting of a form in which the conjunctive form of a verb and the auxiliary verb“ ta ”representing the past are connected”. May be stored in the word DB 21 by distinguishing the past tense of the verb from the original tense of the verb.

  In the word DB updating process of the interactive device 101, the word DB 21 is updated by treating the past tense as a word (word group) different from its original form. Note that all forms other than the past form are handled in the same manner as the interactive device 100 according to the first embodiment, with the original form replaced.

  A specific operation example of such a dialog device 101 will be described with reference to FIG. First, it is assumed that the known information of all the words has been initialized to “NO” in the word DB 21 at the first time. In addition, it is assumed that the model response sentence shown in FIG. 5 is stored in the model response sentence DB 22 together with the keyword. Then, suppose that the user uttered, "I cooked saury at school today, but it was delicious."

  Then, the control unit 10 acquires, as the user utterance sentence, “I cooked the sword fish at school today, but it was delicious.” The word DB 21 is updated as shown in FIG. 14 by the word DB update process. In addition, since the keyword “swordfish” is extracted from the words in the user utterance sentence, the model response sentence acquisition unit 12 reads from the model response sentence DB 22 that “swordfish is delicious if it is greasy with the taste of autumn.” Get response statement.

  The word DB 21 according to the second embodiment treats “delicious” and “delicious” as different words (word groups), and the word group “delicious” exists in the user's utterance sentence. There is no word that is the same as the word "delicious" in the sentence. Therefore, the known information of the word "delicious" remains "NO", and the word of the known information "YES" in the model response sentence is only "swordfish", so a response sentence of "swordfish" is generated. .

  Thereafter, when a user utterance sentence including the words “delicious” and “autumn” is acquired, the word DB update processing changes the known information of these words to “YES”, and the contents of the word DB 21 are shown in FIG. Updated to content. Then, suppose that the user again uttered, "I cooked the sword fish at school today, but it was delicious."

  Then, the control unit 10 obtains “I baked the sword fish at school today, but it was delicious” as the user utterance sentence, and extracted the keyword “Aki sword fish” from the words in the user utterance sentence. 12 acquires an exemplary response sentence from the exemplary response sentence DB 22 stating that "Autumn swordfish is delicious if it is greasy with the taste of autumn." Note that since only words (and word groups) whose known information is already “YES” exist in this user utterance sentence, even if the word DB update process is performed, the word DB 21 does not change, and is shown in FIG. Content remains.

  Then, when the phrase whose known information is “YES” is extracted from the model response sentence, “Akitou is / autumn / delicious” is obtained. In the multiple-word response sentence generation process, the known information of the phrase “autumn” is Since “NO” is set to “NO” (known information of “taste” is “NO” and “autumn” is a modifier of “taste”), the response sentence “Autumn swordfish is delicious” is finally given. Is generated.

  As described above, in the dialogue device 101 according to the second embodiment, the known information of each word (word group) in the word DB 21 becomes “YES” more slowly than in the dialogue device 100. It is possible to show the user how to grow. Therefore, it is possible to make the user feel more growing.

  In the above-described word DB 21, an example in which only the past tense is treated as another word (word group) is shown. However, the present invention is not limited to this, and another form may be treated as another word. For example, a passive form (a verb that can be a passive verb is a "word group consisting of a form in which the auxiliary verb" re, aru "representing the passive and the passive verb" re, aru "), or a negative form (" a verb's intact or adjective " Or, a group of words consisting of a conjunctive form of an adjective verb and a form of conjunctive auxiliary verb "no", which is a concatenation of adjective verbs, ") You may make it memorize | store in the word DB21 as a word different from the original form of the word used. By increasing the form (word group) treated as another word, the speed at which the known information becomes “YES” becomes slower, so that the user can further feel a growing state.

(Embodiment 3)
In the interactive apparatus 101 according to the second embodiment, the speed at which the known information becomes “YES” is reduced by treating the word to be used as another word (word group) according to the used form. Embodiment 3 in which control can be performed more freely will be described.

  The functional configuration of the interactive device 102 according to the third embodiment is the same as the interactive device 100 according to the first embodiment except for the word DB 21. As shown in FIG. 16, the word DB 21 of the interactive device 102 stores, for each word, the number of times the word has been acquired from the user utterance sentence and a predetermined threshold value for treating the word as a known word. . When the number of times is equal to or more than the threshold, the known information of the word is “YES”, and when the number is less than the threshold, the known information of the word is “NO”. As the threshold value, the same value may be set for all the words, or different values may be set according to the words.

  Also, as shown in FIG. 17, in the word DB updating process of the interactive device 100 (FIG. 7), when the determination in step S203 is No, as shown in FIG. , Steps S211 and S212.

  In step S211, the control unit 10 adds 1 to the number of times of the word in the word DB 21 for the word read in step S202. In this step, the control unit 10 functions as a counting unit. Then, it is determined whether or not the number of times after the addition is less than the threshold value stored in the word DB 21 (step S212). If it is less than the threshold value (step S212; Yes), the process proceeds to step S205. If it is equal to or greater than the threshold value (step S212; No), the process proceeds to step S204, and the known information of the word DB 21 for the read word is updated to “YES”.

  When the above-described processing is performed, the timing at which the known information becomes “YES” can be adjusted depending on the word, so that the user can realize growth. For example, if the threshold value of the word of the food name is set to 1, the known information of the name of the food that the user has spoken at least once is “YES”, so the property of “remember the food name immediately” Can be held in the interactive device 102. Also, if the threshold of the name of a sport or the name of a person is set to a large value (for example, 10), the known information of the word becomes “YES” until the user speaks the word more times than the threshold. Therefore, the interactive device 102 can have a property that "the name of a sport or the name of a person is not easily remembered".

  As described above, in the interactive device 102 according to the third embodiment, the timing at which the known information becomes “YES” can be changed, so that the growth can be performed with a higher degree of freedom, and the user can feel the growth. it can. The timing at which the known information becomes “YES” does not have to be set to the same threshold value for all the words, and may be changed for each word type (for example, part of speech) or for each individual word. Is also good. By doing so, the timing at which the known information of each word becomes “YES” can be changed more freely, so that the user can feel further growth.

  In the third embodiment, the number of times and the threshold value of each word are stored in the word DB 21. However, the present invention is not limited to this. For example, the date and time when the word was obtained is recorded in the word DB 21, and a process such as setting the known information to “YES” unless the word is continuously obtained for a predetermined number of days may be performed. Good.

  In the above-described embodiment, the known information in the word DB 21 is updated to “YES” for the words included in the utterance sentence uttered by the user, but the present invention is not limited to this. Regardless of whether or not it is included in the utterance sentence, the known information of each word stored in the word DB 21 may be updated to “YES” according to the time when the user uses the interactive device. In this case, the word DB 21 stores a threshold value of the use time for each word in place of the threshold value of the number of times as shown in FIG. Update to "YES". Thus, even if the user does not readily speak a word, if the time used by the user exceeds a predetermined threshold, the interactive device can respond to the user with a response sentence using such a word. .

  Further, both the threshold of the number of times and the threshold of the use time are stored in the word DB 21, and the number of times included in the user utterance sentence is equal to or greater than the threshold of the number of times or the use time of the user is equal to or greater than the threshold of the use time. Based on this, the known information of the word may be updated to “YES”. By doing so, the timing at which the known information becomes “YES” can be set more freely, so that growth with a higher degree of freedom can be achieved.

  In the above-described embodiment, the word DB 21 is updated by the user utterance sentence obtained by the voice input unit 31 performing speech recognition on the utterance sentence uttered by the user, but the present invention is not limited to this. . For example, a sentence obtained by the voice input unit 31 performing voice recognition of a conversation performed around the interactive devices 100, 101, and 102, such as a voice of another person flowing from a television or a radio, a voice of a user's family or a friend, or the like, is also used. The word DB 21 may be updated.

  In addition, the interactive devices 100, 101, and 102 regard the sentence input as character data in addition to or instead of the voice as a user utterance sentence, update the word DB 21, and generate a response sentence. Or you may. For example, the interactive devices 100, 101, and 102 may regard a sentence composed of characters input by the user from the operation input unit 34 (keyboard, touch panel, etc.) as a user utterance sentence. In this case, the operation input unit 34 functions as a character input unit that receives a character input. When the user inputs a user utterance sentence exclusively from the operation input unit 34 without using voice, the interactive devices 100, 101, and 102 do not need to include the voice input unit 31.

  Further, in the above-described embodiment, the words stored in the word DB 21 are basically nouns, verbs, adjectives, and adjective verbs, but are not limited thereto. For example, the word DB 21 may store an adverb, an intransitive verb, and the like in addition to the above. Conversely, the word DB 21 may not store some of them (for example, adjectives). In this case, a word of the part of speech that is not stored in the word DB 21 is treated as a known word even if the sentence input by the user to the interactive devices 100, 101, and 102 is not included at all.

  Although the above-described embodiment has been described mainly on the premise that Japanese sentences are used, the present invention is not limited to Japanese sentences, and the same concept can be applied to other languages. For example, in the case of English, the word DB 21 may basically store nouns, verbs, and adjectives. Also, in English, “phrase” in the description of the above-described multi-word response sentence generation processing does not exist, but instead of “phrase”, the multi-word response sentence is determined based on the dependency information in “word” units. The known information may be corrected by performing the generation process.

  In the above-described embodiment, the storage unit 20 stores the word DB 21 and the model response sentence DB 22. However, the present invention is not limited to this. At least one of the word DB 21 and the model response sentence DB 22 is stored in an external server or the like, and the word DB 21 or the model response sentence DB 22 stored in the external server or the like is accessed via the communication unit 35. Is also good. In this case, the storage unit 20 does not need to store the DB (the word DB 21 or the model response sentence DB 22) stored by the external server or the like.

  Further, in the above-described embodiment, the interaction devices 100, 101, and 102 include the driving unit 33, the operation input unit 34, and the communication unit 35, but none of them is essential. For example, the interaction devices 100, 101, and 102 need not include the driving unit 33. In this case, instead of "respond with gesture" in step S305 of the response sentence generation process (FIG. 8), a fixed form such as "hmm" or "don't understand" is issued from the voice output unit 32 when words cannot be understood. A response may be made with a response sentence. When the operation input unit 34 is not provided, the control unit 10 may start the interactive processing when the power switch is turned on, and may end the interactive processing when the power switch is turned off.

  Further, the interactive devices 100, 101, and 102 may include display means such as a liquid crystal display, and may respond to the user by displaying a response sentence on the display means instead of voice. The dialogue devices 100, 101, and 102 do not need to include the voice output unit 32 when not using voice to respond to the user.

  Further, the above embodiments can be appropriately combined. For example, by combining Embodiment 2 and Embodiment 3, for the words to be used, the number of times and the threshold value are stored for each of the original form and the utilized form (a word group such as a past form), and the known information is stored for each of the original form and the utilized form May be made to have different timings to make “YES”.

  The functions of the interactive devices 100, 101, and 102 can also be performed by a computer such as a normal PC (Personal Computer). Specifically, in the above-described embodiment, the description has been given on the assumption that the program of the interactive processing performed by the interactive devices 100, 101, and 102 is stored in the ROM of the storage unit 20 in advance. However, the program may be stored in a flexible disk, a compact disc-only memory (CD-ROM), a digital versatile disc (DVD) and a magneto-optical disc (MO), a memory card, a USB (Universal Serial-Readable Memory) of a computer such as a USB (Universal Serial). A computer capable of realizing each of the above-described functions may be configured by storing and distributing the program on a recording medium, reading the program into the computer, and installing the program. Further, a computer capable of realizing each of the above-described functions may be configured by distributing the program via a communication network such as the Internet and reading and installing the program on the computer.

  As described above, the preferred embodiments of the present invention have been described. However, the present invention is not limited to the specific embodiments, and the present invention includes the inventions described in the claims and equivalents thereof. It is. Hereinafter, the invention described in the claims of the present application is additionally described.

(Appendix 1)
Parsing means for parsing a user utterance sentence which is an utterance sentence input by the user;
An exemplary response sentence obtaining unit that obtains an exemplary response sentence that is an exemplary response sentence according to a syntax analysis result of the user utterance sentence by the syntax analysis unit;
According to a predetermined condition, word use availability determination means for determining whether or not each word included in the model response sentence acquired by the model response sentence acquisition means is usable or unusable,
Response sentence generation means for generating a response sentence to the user utterance sentence by deleting a word determined to be unusable by the word use availability determination means from the model response sentence acquired by the model response sentence acquisition means,
Response means for responding to the user using the response sentence generated by the response sentence generation means,
An interactive device comprising:

(Appendix 2)
The predetermined condition includes a condition that a word included in the model response sentence is a word corresponding to a word included in the parsed user utterance sentence,
The word availability determining unit determines that a word corresponding to a word included in the parsed user utterance sentence among the words included in the model response sentence is usable, and the syntax analysis is performed. Judge that a word corresponding to a word not included in the user utterance sentence is not usable,
The interactive device according to supplementary note 1.

(Appendix 3)
The predetermined condition is a word used by a word included in the model response sentence, and a word group in a form in which a predetermined auxiliary form is connected to a predetermined usage form of the word to be used is included in the model response sentence. In the case, the condition includes whether or not the same word group as the word group is included in the parsed user utterance sentence,
When the word included in the model response sentence is a word to be used, the word use availability determination unit may determine that the model response sentence includes the same word group as the word group including the word in the user utterance sentence. Among the words included in the word group, it is determined that the word corresponding to the word group is usable. If the word group is not included in the user utterance sentence, the word among the words included in the model response sentence is determined. Judge that the word corresponding to the group cannot be used,
The interactive device according to supplementary note 2.

(Appendix 4)
Further, a counting means for counting the number of times each word included in the user utterance sentence is input by the user,
The predetermined condition further includes a condition that the number of times counted by the counting means is equal to or more than a predetermined threshold,
The word availability determining unit determines that the number of times counted by the counting unit is a word corresponding to a word having a predetermined threshold or more, and determines that the word corresponding to the word having the number of times less than the threshold is unusable. judge,
The interactive device according to Supplementary Note 2 or 3.

(Appendix 5)
The threshold comprises a plurality of thresholds set for each of a plurality of different words,
The interactive device according to supplementary note 4.

(Appendix 6)
The word use availability determination means, when a word determined to be unusable exists in the model response sentence, a word corresponding to a modification source word that modifies the word determined to be unusable is included in the user utterance sentence. Even if it is included, it is determined that the word of the modifier is also unusable,
6. The interactive device according to any one of supplementary notes 2 to 5.

(Appendix 7)
Furthermore, it comprises a voice input means for voice recognition to obtain the user utterance sentence,
The syntax analysis unit performs syntax analysis of the user utterance sentence obtained by the voice input unit performing voice recognition,
7. The interactive device according to any one of supplementary notes 1 to 6.

(Appendix 8)
Furthermore, it has voice output means for outputting voice by voice synthesis,
The response means responds to the user by outputting the response sentence by voice by the voice output means,
8. The interactive device according to any one of supplementary notes 1 to 7.

(Appendix 9)
Furthermore, a character input means for receiving an input of a character from a user is provided,
The syntax analysis unit analyzes a sentence composed of characters received by the character input unit as the user utterance sentence,
The interactive device according to any one of supplementary notes 1 to 8.

(Appendix 10)
Furthermore, a display unit for displaying characters is provided,
The response unit responds to the user by displaying the response sentence on the display unit.
10. The interactive device according to any one of supplementary notes 1 to 9.

(Appendix 11)
Further, a driving unit for performing an operation indicating that the user utterance sentence cannot be understood,
The response unit, when the response sentence generation unit cannot generate the response sentence, responds to the user by an operation of the driving unit.
11. The interactive device according to any one of supplementary notes 1 to 10.

(Appendix 12)
An interactive device control method for controlling an interactive device,
A parsing step of parsing a user utterance sentence that is an utterance sentence input by the user;
An exemplary response sentence obtaining step of obtaining an exemplary response sentence that is an exemplary response sentence according to the result of the syntax analysis of the user utterance sentence by the syntax analysis step;
According to a predetermined condition, a word use determination step of determining whether or not each word included in the model response sentence acquired in the model response sentence acquisition step is usable or unusable,
A response sentence generation step of generating a response sentence to the user utterance sentence by deleting a word determined to be unusable in the word use availability determination step from the model response sentence acquired in the model response sentence acquisition step;
A response step of responding using the response sentence generated in the response sentence generation step,
An interactive device control method including:

(Appendix 13)
On the computer of the interactive device,
A parsing step of parsing a user utterance sentence that is an utterance sentence input by the user;
An exemplary response sentence obtaining step of obtaining an exemplary response sentence that is an exemplary response sentence according to a result of the syntax analysis of the user utterance sentence by the syntax analysis step;
According to a predetermined condition, a word use determination step of determining whether or not each word included in the model response sentence acquired in the model response sentence acquisition step is usable or unusable,
A response sentence generation step of generating a response sentence to the user utterance sentence by deleting a word determined to be unusable in the word use availability determination step from the model response sentence acquired in the model response sentence acquisition step, and ,
A response step of responding using the response sentence generated in the response sentence generation step,
The program to execute.

DESCRIPTION OF SYMBOLS 10 ... Control part, 11 ... Syntax analysis part, 12 ... Model response sentence acquisition part, 13 ... Word availability determination part, 14 ... Response sentence generation part, 15 ... Response part, 20 ... Storage part, 21 ... Word DB, 22 ... model response sentence DB, 31 ... voice input unit, 32 ... voice output unit, 33 ... drive unit, 34 ... operation input unit, 35 ... communication unit, 100, 101, 102 ... dialogue device

Claims (13)

Parsing means for parsing a user utterance sentence which is an utterance sentence input by the user;
An exemplary response sentence obtaining unit that obtains an exemplary response sentence that is an exemplary response sentence according to a syntax analysis result of the user utterance sentence by the syntax analysis unit;
According to a predetermined condition, word use availability determination means for determining whether or not each word included in the model response sentence acquired by the model response sentence acquisition means is usable or unusable,
Response sentence generation means for generating a response sentence to the user utterance sentence by deleting a word determined to be unusable by the word use availability determination means from the model response sentence acquired by the model response sentence acquisition means,
Response means for responding to the user using the response sentence generated by the response sentence generation means,
An interactive device comprising: The predetermined condition includes a condition that a word included in the model response sentence is a word corresponding to a word included in the parsed user utterance sentence,
The word availability determining unit determines that a word corresponding to a word included in the parsed user utterance sentence among the words included in the model response sentence is usable, and the syntax analysis is performed. Judge that a word corresponding to a word not included in the user utterance sentence is not usable,
The interactive device according to claim 1. The predetermined condition is a word used by a word included in the model response sentence, and a word group in a form in which a predetermined auxiliary form is connected to a predetermined usage form of the word to be used is included in the model response sentence. In the case, the condition includes whether or not the same word group as the word group is included in the parsed user utterance sentence,
If the word included in the model response sentence is a word to be utilized, the word use availability determination unit may determine the model response sentence if the same word group as the word group including the word is included in the user utterance sentence. Among the words included in the word group, it is determined that the word corresponding to the word group is usable. If the word group is not included in the user utterance sentence, the word among the words included in the model response sentence is determined. Judge that the word corresponding to the group cannot be used,
The interactive device according to claim 2. Further, a counting means for counting the number of times each word included in the user utterance sentence is input by the user,
The predetermined condition further includes a condition that the number of times counted by the counting means is equal to or more than a predetermined threshold,
The word availability determining unit determines that the number of times counted by the counting unit is a word corresponding to a word that is equal to or greater than a predetermined threshold, and determines that the word corresponding to a word whose number is less than the threshold is unusable. judge,
The interactive device according to claim 2. The threshold comprises a plurality of thresholds set for each of a plurality of different words,
The interactive device according to claim 4. The word use availability determination means, when a word determined to be unusable exists in the model response sentence, a word corresponding to a modification source word that modifies the word determined to be unusable is included in the user utterance sentence. Even if it is included, it is determined that the word of the modifier is also unusable,
The interactive device according to any one of claims 2 to 5. Furthermore, it comprises a voice input means for voice recognition to obtain the user utterance sentence,
The syntax analysis unit performs syntax analysis of the user utterance sentence obtained by the voice input unit performing voice recognition,
The interactive device according to claim 1. Furthermore, it has voice output means for outputting voice by voice synthesis,
The response means responds to the user by outputting the response sentence by voice by the voice output means,
The interactive device according to claim 1. Furthermore, a character input means for receiving an input of a character from a user is provided,
The syntax analysis unit analyzes a sentence composed of characters received by the character input unit as the user utterance sentence,
The interactive device according to claim 1. Furthermore, a display unit for displaying characters is provided,
The response unit responds to the user by displaying the response sentence on the display unit.
The interactive device according to claim 1. Further, a driving unit for performing an operation indicating that the user utterance sentence cannot be understood,
The response unit, when the response sentence generation unit cannot generate the response sentence, responds to the user by an operation of the driving unit.
The interactive device according to claim 1. An interactive device control method for controlling an interactive device,
A parsing step of parsing a user utterance sentence that is an utterance sentence input by the user;
An exemplary response sentence obtaining step of obtaining an exemplary response sentence that is an exemplary response sentence according to the result of the syntax analysis of the user utterance sentence by the syntax analysis step;
According to a predetermined condition, a word use determination step of determining whether or not each word included in the model response sentence acquired in the model response sentence acquisition step is usable or unusable,
A response sentence generation step of generating a response sentence to the user utterance sentence by deleting a word determined to be unusable in the word use availability determination step from the model response sentence acquired in the model response sentence acquisition step;
A response step of responding using the response sentence generated in the response sentence generation step,
An interactive device control method including: On the computer of the interactive device,
A parsing step of parsing a user utterance sentence that is an utterance sentence input by the user;
An exemplary response sentence obtaining step of obtaining an exemplary response sentence that is an exemplary response sentence according to a result of the syntax analysis of the user utterance sentence by the syntax analysis step;
According to a predetermined condition, a word use determination step of determining whether or not each word included in the model response sentence acquired in the model response sentence acquisition step is usable or unusable,
A response sentence generation step of generating a response sentence to the user utterance sentence by deleting a word determined to be unusable in the word use availability determination step from the model response sentence acquired in the model response sentence acquisition step, and ,
A response step of responding using the response sentence generated in the response sentence generation step,
The program to execute.

Images