JP5988501B2 - Dialog action output device, method, and program, and dialog system and method - Google Patents

Description

  The present invention relates to an interactive action output device, method, and program, and an interactive system and method.

  Various researches have been conducted on character characteristics of dialogue systems, and some have been commercialized. For example, there are many products on the market that synthesize speech from female, male, various ages, character voices, etc. In addition, software that realizes smooth communication by showing an animation agent on a robot or a screen has been proposed (see Non-Patent Document 1, for example).

  In addition to multi-modal speech such as voice and face images, research has also been conducted to give linguistic character characteristics to dialog systems (for example, see Non-Patent Document 2).

Shigeki Hiyama et al., “Development project report of anthropomorphic spoken dialogue agent basic software”, IEICE technical report, NLC, language understanding and communication, pp 73-78, 2003. Kazue Kaneko, Tsuyoshi Yagisawa, Satoshi Fujita, “Sentence Generation System that Reflects Gender and Age of Speaker”, Information Processing Society of Japan Research Report, Natural Language Processing Study Group Report, Vol 99, No. 114, pp129-136, 1996.

  As described above, in the previous research, in the fields of multimodal and sentence generation, research has been conducted to give the dialog system (dialog agent) character characteristics. However, the character of the dialogue system is greatly influenced by the impression of the personality based on the dialogue act indicating the intention of the dialogue system. For example, a dialogue system that actively self-disclosures in a dialogue receives an impression of an outgoing character. On the other hand, a dialogue system with many fillers gives a negative impression.

  In previous studies, there are things that change the impression of the nature of the dialogue system given to the user by changing the appearance and behavior of the dialogue system. However, as described above, the character of the interactive system felt by the user is influenced not only by facial expressions, gestures and wording while the interactive system makes one utterance, but also by the personality impression based on the interactive action. For this reason, it is difficult to change the impression of personality based on the dialogue action as described above only by changes in appearance and behavior, and it may not be possible to express clear character characteristics.

  The present invention has been made in view of the above problems, and an interactive action output device, method, and program for outputting a type of interactive action that can clarify character characteristics, and an interactive dialog that clarifies character characteristics. It is an object to provide a system and method.

In order to achieve the above object, the dialogue action output device according to the first invention has a plurality of different dialogue action types indicating the intention of the utterance, and has the certainty as the type of dialogue action used to generate the utterance. The value indicating the likelihood of at least some of the types of interactive actions assigned to each of the values to be indicated is determined according to a rule preliminarily determined according to the character assumed by the speaker of the generated utterance and the probability a dialog act processing unit for changing, based on the value indicating the value indicating the probability that a processing for outputting the type of the plurality of dialog act granted to each of the plurality of by the dialogue act processing unit If the value indicating the likelihood given to the type of each of the dialog act is changed granted value indicating the probability of the changed type each of the plurality of dialogue act, used to generate the speech It is configured to include a, and an output unit that performs processing for outputting the type of the plurality of dialog acts.

According to the interactive action output device according to the first invention, the interactive action processing unit has a plurality of different interactive action types indicating the intention of the utterance, and the probability as the interactive action type used for generating the utterance is determined. The value indicating the likelihood of at least some of the types of interactive actions assigned to each of the values to be indicated is determined according to a rule preliminarily determined according to the character assumed by the speaker of the generated utterance and the probability It changes based on the value which shows . Then, the output unit, a processing value indicating the likelihood outputs the type of the plurality of dialog act granted to each probability after the change when the value is changed by the dialogue act processor the granted to type each value of the plurality of dialog acts showing performs processing for outputting the type of the plurality of dialog acts used to generate speech.

  As a result, it is possible to output the type of interactive action that can clarify the character characteristics.

  Further, the predetermined rule is determined to increase or decrease the value for each type of dialogue action based on the correlation between the index characterizing the character assumed for the speaker and each type of dialogue action. It can be a rule. As a result, it is possible to output the type of dialogue action that can emphasize the character assumed for the speaker and clarify the character.

The dialogue system according to the second aspect of the present invention includes an utterance analysis unit that analyzes a user's utterance and analyzes a type of user interaction act indicating the user's utterance intention, and a system interaction act that indicates the intention of the system utterance. As a plurality of candidates, a plurality of different types of system interaction actions, each of which is given a value indicating the certainty as the type of system interaction action with respect to the type of user interaction action analyzed by the speech analysis unit, are output. a plurality of dialogue action candidate output unit, a plurality of candidate types of systems dialog act indicating the intent of the utterance of the system, the value that indicates the likelihood that is output from the dialogue act candidate output unit is applied to each of the Le that of a value indicative of at least a portion of the probability of the type of system dialogue act, determined in advance in accordance with the nature envisioned for the system Change based on the value indicating the Le and the likelihood, the type of each of the values indicating the likelihood that a process of outputting the type of said plurality of systems dialog act granted to each of the plurality of dialogue act When the value indicating the certainty given to the change is changed, the value indicating the probability after the change is assigned to each of the plurality of types of interactive action, and the plurality of types of the system interactive action used for generating the utterance Utterance of the system based on an interactive action output unit that performs a process of outputting the system, and a plurality of system interactive action types that are output by the interactive action output unit and that are assigned values indicating the certainty. An utterance generation unit that generates and outputs the utterance.

  Thereby, the character property of a dialog system can be clarified.

A dialogue act output method according to a third invention is a dialogue act output method in a dialogue act output device including a dialogue act processing unit and an output unit, wherein the dialogue act processing unit indicates an intention of utterance. different of the type of a plurality of dialog acts, a value indicating at least a portion of the probability of the type of certainty multiple interactions acts value is assigned to each illustrating a as type of dialogue act used to generate speech and a step of changing, based on the value indicating the predetermined rule and the likelihood in accordance with the nature that is assumed speaker speech generated, the output unit, the value indicating the likelihood that each a process for outputting granted type of the plurality of dialog acts, if the value indicating the likelihood given to the type of each of the plurality of dialog act by the dialogue act processing unit is changed Is a method comprising granted value indicating the probability of the changed type each of the plurality of dialogue act, and performing processing for outputting the type of the plurality of dialog acts used to generate speech, the.

A dialogue method according to a fourth invention is a dialogue method in a dialogue system including an utterance analysis unit, a dialogue action candidate output unit, a dialogue action output unit, and an utterance generation unit, wherein the utterance analysis unit includes: Analyzing a user's utterance and analyzing a user interaction action type indicating the user's utterance intention; and the dialog action candidate output unit includes a plurality of system interaction action types indicating the utterance intention of the system . As a candidate, outputting a plurality of different types of system interaction actions each assigned a value indicating the probability as the type of system interaction action with respect to the type of user interaction action analyzed by the utterance analysis unit; dialog act output unit, a plurality of candidate types of systems dialog act indicating the intent of the utterance of the system, output from the dialogue act candidate output unit A value indicating the likelihood of at least some types of the likelihood of multiple systems dialog acts a value is assigned to each showing, the rules and the likelihood determined in advance in accordance with the nature envisioned for the system change based on the value indicating the value indicating the probability that a processing for outputting the type of said plurality of systems dialog act granted to each likelihood given to the type of each of the plurality of dialogue act When the value indicating the change is changed , a value indicating the certainty after the change is assigned to each of the plurality of interaction action types, and the process of outputting the plurality of system interaction action types used for utterance generation is performed. a step, the speech generating unit, the type of the dialogue act output unit output by the value indicating the likelihood is assigned the multiple systems dialogue act, Zui by a method comprising the steps of: generating and outputting an utterance of said system.

  Moreover, the interactive action output program which concerns on 5th invention is a program for functioning a computer as each part which comprises said interactive action output device.

As described above, according to the interactive action output device, method, and program of the present invention, a value indicating the certainty as the type of interactive action used for generating an utterance is assumed for the utterer of the generated utterance. that for outputting a type of variable puff dialogue action based on a predetermined rule and the values assigned to the type of dialogue act in accordance with the nature, the output type of dialog acts that can clarify the character of can do.

Further, according to the dialogue system and method of the present invention, the value indicating the certainty as the type of dialogue action used for generating the system utterance is determined based on the rule and the type of dialogue action determined in advance according to the character assumed for the system. based on the type of dialogue act that strange owl based on the assigned value in order to generate a speech system, it is possible to clarify the character of the system.

It is a block diagram which shows the structure of the dialogue system which concerns on 1st Embodiment. It is a figure which shows an example of a dialogue action set. It is a block diagram which shows the structure of the dialogue action output part which concerns on 1st Embodiment. It is a figure which shows an example of the type of a system interaction action type. It is a figure which shows the personality characteristic for every speaker, and the appearance rate of the type of dialogue action. It is a figure which shows an example of dialogue action character characteristic correlation data. It is a figure which shows an example of the type of the system interaction act characterized by the character in 1st Embodiment. It is a flowchart which shows the content of the dialogue process in 1st Embodiment. It is a flowchart which shows the content of the dialogue action output process in 1st Embodiment. It is a block diagram which shows the structure of the dialogue action output part which concerns on 2nd Embodiment. It is a figure which shows an example of the type of the system interaction act characterized by the character in 2nd Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<First Embodiment>
A dialogue system 10 according to the first embodiment includes a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (ROM) storing a program for executing dialogue processing including dialogue action output processing described later. Read only memory). When the CPU reads and executes the program from the ROM that is an internal storage device, the computer functions as the interactive system 10.

  The dialogue system 10 can be functionally represented by a configuration including an utterance analysis unit 20, a dialogue action candidate output unit 30, a dialogue action output unit 40, and an utterance generation unit 50, as shown in FIG. The dialogue action output unit 40 is an example of the dialogue action output device of the present invention.

  The utterance analysis unit 20 receives a user's utterance sentence (text data) in the dialogue. Note that a voice recognition unit may be provided to accept a user's speech as voice data, and the voice recognition unit may convert the voice data into text data. The utterance analysis unit 20 analyzes an input user utterance sentence (hereinafter referred to as “user utterance”) and a type of dialogue action indicating the user's intention in the user utterance (hereinafter referred to as “type of user dialogue action”). Is output.

  Here, the type of dialogue action can be selected from, for example, a dialogue action set as shown in FIG. In the example of FIG. 2, there are seven sub-categories of “actual”, “experience”, “custom”, “evaluation”, “desire”, “plan”, and “other” as the type of dialogue action “self-disclosure”. (Sub) is defined, and the subcategory “evaluation” is further classified into three different subcategories: “positive”, “negative”, and “neutral”. The dialogue action set is not limited to the example of FIG. 2, and various dialogue action sets can be used.

  Specifically, the utterance analysis unit 20 performs morphological analysis on the user utterance, and extracts a feature such as a word n-gram from the user utterance using the morphological analysis result. The user interaction action is output based on the classification model learned using the learning data that paired the feature extracted from the utterance and the correct answer of the dialogue action type of the utterance and the feature extracted from the user utterance. To do. In addition, the output method of a user interaction act is not limited to this method.

  The dialogue action candidate output unit 30 accepts the user dialogue action output from the utterance analysis unit 20 as an input, and as a candidate for the type of system dialogue action for the user dialogue action (hereinafter referred to as “type of system dialogue action”), Outputs different types of system interaction actions. The dialogue action candidate output unit 30 may output, for example, a system dialogue action type candidate corresponding to the user dialogue action according to a predetermined rule, a model learned using a dialogue sentence for learning, or the like. May be used to output system interaction action type candidates.

  The system interaction action type candidates output here are obtained by assigning a value indicating the probability of the system interaction action type to the user interaction action type for each system interaction action type candidate. It may be present or may not be assigned a value. The value may be, for example, a score such as an appearance probability value for each type of system interaction action with respect to the type of user interaction action, or may be a rank based on such a score. Further, the same score, ranking, and other values may be assigned to all system interaction action type candidates.

  The dialogue action output unit 40 calculates a value given to each of the system dialogue action type candidates based on the personality feature set according to the character of the dialogue system 10 (the dialogue agent expressed by the dialogue system 10). change. When no value is assigned to the system interaction action type candidate, a value based on the personality feature set according to the character of the interaction system 10 is assigned to each of the system interaction action type candidates.

  As shown in FIG. 3, the dialogue act output unit 40 can be represented by a configuration including an input unit 41, a dialogue act processing unit 42, a normalization unit 43, and an output unit 44.

  The input unit 41 receives as input the candidate set of system interaction action types output from the interaction action candidate output unit 30 and passes it to the interaction action processing unit 42. For example, a candidate set of system interaction action types as shown in FIG. 4 is input. In the example of FIG. 4, each candidate (type name of dialogue action) and the score given to the candidate are written together.

  The dialogue action processing unit 42 refers to the dialogue act personality characteristic correlation data 61 and the dialogue act emphasis rule 62 for each of the candidates included in the candidate set of system dialogue act types passed from the input unit 41, and character characteristics And increasing or decreasing the value for each of the system interaction type candidates based on the correlation between and the interaction type. For example, if there is a positive correlation between the personality trait and the type of dialogue, increase the value of that dialogue type, and if there is a negative correlation between the personality feature and the type of dialogue, or if the correlation is If not, the interactive action type value can be reduced.

  Specifically, the dialogue action personality feature correlation data 61 is prepared as follows. First, the value of the personality feature is measured for each of a plurality of speakers. Here, it is called Big-Five that can measure the strength of each personality feature by dividing the personality into five axes representing the personality features of extroversion, attachment, control, emotionality, and playability. Use the technique. This method allows a speaker to select a value of 1 to 5 indicating the degree of 50 question items (for example, “a person who meets the person would like it most”). A score of ˜50 can be given. In addition, you may use not only Big-Five but the other parameter | index showing a person's character.

  The type of dialogue action is manually assigned to all of the utterances included in the dialogue between the plurality of speakers. Then, for each speaker, the appearance ratio of the utterance for each type of dialogue action with respect to the total number of utterances of the corresponding speaker during the dialogue is calculated. For example, FIG. 5 shows an example of the extroversion value of each speaker and the appearance ratio of the speech whose greeting action is “greeting” when 37 speakers and 1260 dialogues are recorded. Based on the tabulation result as shown in FIG. 5, the correlation value between the value of each personality feature and the appearance ratio of the dialogue action type is calculated. Such a correlation value is calculated for each personality feature and for each type of dialogue action. FIG. 6 shows an example of the correlation value between each personality feature value and the appearance rate of the dialogue action type, that is, dialogue action personality feature correlation data 61.

  In addition, a dialogue act emphasis rule 62 that determines which personality feature to emphasize according to the character of the system is determined in advance. For example, in the case of setting the character property with strong extroversion, the correlation value between the extroversion value and the appearance rate of the type of dialogue act is a predetermined threshold (for example, 0.2 ) Raise the score given to the above interaction action type candidates (for example, 10 times), and the correlation value between the extroversion value and the appearance ratio of the interaction action is a predetermined threshold (for example, −0. 2) A rule for lowering (for example, 1/10) a score given to a candidate for the following interactive action type can be set.

  In addition, when there is no correlation between the personality feature that emphasizes the personality set according to the character of the system and the appearance ratio of the type of dialogue action (for example, the correlation value is 0.1 to −0.1), the dialogue is performed. Rules may be established that lower the score of the type of action.

  In addition, when ranks are given to the input system dialogue action candidates, the correlation value between the value of the specific personality feature and the appearance rate of the dialogue action type is equal to or higher than a predetermined threshold value. It is possible to set a rule that raises the rank of candidates and lowers the rank of candidates for interactive action types that are below a predetermined threshold. In addition, if the input system dialogue action candidate is not given a value such as a score or a ranking, the score or the score in descending order of the correlation value between the value of the specific personality feature and the appearance rate of the type of dialogue action. Rules that give rankings can be established.

  In the example of the dialogue act personality characteristic correlation data 61 shown in FIG. 6, the types of dialogue act having a correlation value with an extroversion value of 0.2 or more are information provision, self-disclosure_facts, self-disclosure_experience, self-disclosure. _Evaluation- and self-disclosure_scheduled. Therefore, the dialogue action processing unit 42, for example, increases the score given to these dialogue action type candidates by ten times. On the other hand, the types of dialogue actions whose correlation value with the extroversion value is −0.2 or less are greetings, companions, questions_request for information provision, questions_habits, questions_desires, non-sympathy / disagreement, thanks, Filler and exclamation. Accordingly, the dialogue action processing unit 42, for example, multiplies the score given to these dialogue action type candidates by 1/10.

  The dialogue action processing unit 42 outputs a candidate set of system dialogue action types in which the scores are changed according to the dialogue act personality characteristic correlation data 61 and the dialogue act emphasis rules.

  The normalization unit 43 receives as input the candidate set of system interaction action types output from the interaction action processing unit 42, normalizes the changed score to be 1, and outputs the result. FIG. 7 shows an example in which the dialogue action processing unit 42 and the normalization unit 43 perform the process for the candidate set of system dialogue action types in FIG. In the example of FIG. 7, self-disclosure_facts, information provision, etc. often performed by people with high extroversion are at the top of the ranking.

  The output unit 44 outputs the normalized type of system interaction action output from the normalization unit 43 as the type of system interaction action characterized by personality.

  The utterance generation unit 50 generates and outputs a system utterance based on the type of system interaction action characterized by the character output from the interaction action output unit 40. A conventionally known method can be used to generate an utterance.

  Next, the operation of the dialogue system 10 according to the first embodiment will be described. When a user utterance is input to the dialogue system 10, the dialogue processing shown in FIG. 8 is executed in the dialogue system 10.

  In step 100 of the dialogue process shown in FIG. 8, the utterance analysis unit 20 analyzes the input user utterance and outputs the type of user interaction act indicating the user's intention in the user utterance.

  Next, in step 200, the dialogue action candidate output unit 30 accepts the type of user interaction action output from the utterance analysis unit 20 as an input, and outputs at least one system interaction action type candidate for the user interaction action. .

  Next, in step 300, the dialogue action output unit 40 executes the dialogue action output process shown in FIG.

  In step 301 of the dialogue action output process shown in FIG. 9, the input unit 41 accepts a candidate set of system dialogue action types output from the dialogue action candidate output unit 30 as an input, and passes it to the dialogue act processing unit 42.

  Next, in step 302, the dialogue action processing unit 42, for each candidate included in the candidate set of system dialogue action types passed from the input unit 41, the dialogue act personality feature correlation data 61 and the dialogue act emphasis rule. Referring to 62, based on the correlation between the personality characteristics and the interaction action type, the value for each of the system interaction action type candidates is changed and output to be increased or decreased.

  Next, in step 303, the normalization unit 43 accepts as input a candidate set of system interaction action types with changed scores output from the interaction action processing unit 42, and the total of the changed scores becomes 1. Normalize so that

  Next, in step 304, the output unit 44 outputs the candidate of the system interaction action type with the normalized score output from the normalization unit 43 as the type of the system interaction action characterized by the personality. The process returns to the dialogue process shown in FIG.

  Next, in step 400 of the dialogue process shown in FIG. 8, the utterance generation unit 50 generates a system utterance based on the type of system dialogue action characterized by the character output from the dialogue action output unit 40. Output and finish the interactive process.

  As described above, according to the interactive action output unit of the interactive system according to the first embodiment, the value for each type of system interactive action indicating the intention of system utterance is changed according to the character of the system. Thus, it is possible to output the type of interactive action that can clarify the character characteristics. Therefore, in the interactive system, the character of the system can be clarified. Thereby, a dialogue system that each user likes can be constructed, dialogue can be smoothly performed, intellectual activities can be promoted, and individual systems can have different personalities, thereby enhancing humanity.

<Second Embodiment>
Next, a second embodiment will be described. As shown in FIG. 1, the dialogue system 210 according to the second embodiment differs from the dialogue system 10 according to the first embodiment only in the dialogue action output unit. The interactive action output unit 240 in the embodiment will be described. In addition, about the part similar to the interactive action output part 40 in the interactive system 10 which concerns on 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  As shown in FIG. 10, the dialogue action output unit 240 can be represented by a configuration including an input unit 41, a dialogue action processing unit 242, a normalization unit 43, and an output unit 44.

  Unlike the dialogue action processing unit 42 in the first embodiment, the dialogue action processing unit 242 directly follows the dialogue action emphasis rule 262 and does not depend on the correlation between the personality characteristics and the type of dialogue action. Change the candidate value of. As the dialogue action emphasis rule 262, for example, a rule for randomly changing the score can be determined according to a distribution manually set according to the character of the system. In addition, a rule for changing the score of the interactive action type according to the character of the system can be determined based on a predetermined relationship between the personality and the interactive action type capable of emphasizing the personality. . For example, personality is based on the relationship that the type of dialogue act is often “self-disclosure” in the active personality, and the relationship that the type of dialogue act is “filler” in the passive character. Rules that increase or decrease the value of the type of interactive action related to the. More specifically, if you want to give the system the character nature of “speaking only about yourself,” the score of the system interaction action type that is “self-disclosure” among the candidate system interaction action types. A rule of raising (for example, 10 times) can be determined.

  FIG. 11 shows an example in which the dialogue action processing unit 242 and the normalization unit 43 perform the processing for the candidate set of system dialogue action types in FIG. In the example of FIG. 11, the score of the dialogue action type “self-disclosure” is normalized by 10 times. It can be seen that the ranking of “self-disclosure” is at the top.

  The operation of the dialogue system 210 according to the second embodiment is the dialogue action emphasis directly in step 302 of the dialogue action output process in the first embodiment regardless of the correlation between the personality characteristics and the type of the dialogue action. Since the point of changing the score of the system interaction action type candidate according to the rule 262 is only different from the operation of the interaction system 10 according to the first embodiment, detailed description thereof is omitted.

  As described above, the dialog system according to the second embodiment can achieve the same effects as those of the first embodiment.

  The present invention is not limited to the above embodiment, and various modifications and applications can be made without departing from the gist of the present invention.

  In the above embodiment, the case where the dialogue action output unit 40 (or 240) is incorporated in the dialogue system 10 (or 210) has been described, but only the configuration of the dialogue action output unit 40 (or 240). A single device may be configured. In this case, any type of system interaction action, user interaction action type, or interaction action type may be input to the interaction action output device. For example, a user interaction act type candidate set is accepted as an input, and based on the user's preset accuracy, the value assigned to each of the candidates is changed to characterize the user's personality The type may be output. The type of user interaction action output in this way can be used in, for example, a system that predicts the user's next utterance.

  Further, in the present specification, the embodiment in which the program is preinstalled has been described. However, the program stored in an external storage device or recording medium is read as needed, and is downloaded and executed via the Internet. You may make it do. In addition, the program can be provided by being stored in a computer-readable recording medium.

DESCRIPTION OF SYMBOLS 10,210 Dialog system 20 Speech analysis part 30 Dialogue action candidate output part 40,240 Dialogue action output part 41 Input part 42,242 Dialogue action processing part 43 Normalization part 44 Output part 50 Speech generation part 61 Dialogue action character characteristic correlation data 62,262 Dialogue action emphasis rules

Claims (6)

A plurality of different types of interactive actions that indicate the intention of the utterance, wherein at least some of the types of interactive actions that are each given a value indicating the certainty as the type of interactive action used to generate the utterance a value indicating the likelihood, the dialog act processing unit for changing, based on the value indicating the predetermined rule and the likelihood in accordance with the nature that is assumed speaker speech generated,
The value indicating the probability that a processing for outputting the type of the plurality of dialog act granted to each value indicating the dialogue act processing unit by likelihood given to the type of each of the plurality of dialogue act A value indicating the certainty after the change is assigned to each of the plurality of interaction action types, and an output unit that performs processing to output the plurality of interaction action types used for generating an utterance ; ,
Interactive action output device including: The predetermined rule is increased or decreased based on a correlation between an index characterizing the personality assumed for the speaker and each type of dialogue action, and the value indicating the likelihood for each type of dialogue action. The interactive action output device according to claim 1, wherein the rule is defined as a rule. An utterance analysis unit that analyzes the user's utterance and analyzes the type of user interaction act indicating the intention of the user's utterance;
As a plurality of candidates for the type of system interaction act indicating the intention of the system utterance, a value indicating the probability as the type of system interaction action with respect to the type of user interaction action analyzed by the utterance analysis unit is given to each A dialogue action candidate output unit for outputting different types of system dialogue actions;
As a plurality of system interaction action type candidates indicating the utterance intention of the system , at least one of a plurality of system interaction action types each having a value indicating the certainty output from the interaction action candidate output unit is provided. some of the values indicating the likelihood, and change based on the value indicating the predetermined rule and the likelihood in accordance with the nature envisaged in the system, a value indicating the likelihood is applied to each said A process of outputting a plurality of system interaction action types, and when the value indicating the certainty assigned to each of the plurality of interaction action types is changed, the value indicating the probability after the change is the plurality An interactive action output unit that performs processing of outputting the plurality of system interactive action types used to generate an utterance, which is assigned to each interactive action type;
The output by the dialogue act output unit, and the likelihood of the plurality of types of system dialog acts the value was assigned indicating the, on the basis of, speech generator for generating and outputting an utterance of said system,
Interactive system including A dialogue action output method in a dialogue action output device including a dialogue action processing unit and an output unit,
The interactive action processing unit is a plurality of different interactive action types indicating intentions of utterances, and a plurality of interactive action types each having a value indicating the certainty as a type of interactive action used for generating the utterances. a step type a value indicating the likelihood of at least some, to change based on the value indicating the predetermined rule and the likelihood in accordance with the nature that is assumed speaker speech generated,
The output unit, the value indicating the probability that a processing for outputting the type of the plurality of dialog act granted to each were granted to the type of each of the plurality of dialog act by the dialogue act processing unit When the value indicating the certainty is changed, the value indicating the certainty after the change is assigned to each of the plurality of interactive action types, and the plurality of interactive action types used for generating the utterance are output. The steps of
Dialogue action output method including A dialogue method in a dialogue system including an utterance analysis unit, a dialogue action candidate output unit, a dialogue action output unit, and an utterance generation unit,
The utterance analysis unit analyzes the user's utterance and analyzes the type of user interaction act indicating the intention of the user's utterance;
The dialog action candidate output unit has a plurality of system interaction action type candidates indicating the intention of the system utterance, and the probability of the system interaction action type with respect to the type of user interaction action analyzed by the utterance analysis unit. Outputting a plurality of different types of system interaction actions each having a value indicated;
As the plurality of candidates for the type of system interaction action indicating the intention of the system utterance , the interaction action output unit has a plurality of values each given a value indicating the likelihood output from the interaction action candidate output unit a value indicative of at least a portion of the probability of the type of system dialogue act to change based on the value indicating the predetermined rule and the likelihood in accordance with the nature envisaged in the system, a value indicating the likelihood there a process of outputting the type of said plurality of systems dialog act granted to each probability after the change, if you change the value that indicates the likelihood given to the type of each of the plurality of dialogue act the value indicated is applied to the type of each of the plurality of dialog acts, performing a process for outputting the type of the plurality of system dialogue act used to generate speech ,
The step of the speech generating unit, the dialog act is output by the output unit, the likelihood of the plurality of types of system dialog acts the value was assigned indicating the, on the basis, and generates and outputs an utterance of said system When,
Interactive method including An interactive action output program for causing a computer to function as each part constituting the interactive action output device according to claim 1.

Images