JP7176333B2 - Dialogue device, dialogue method and dialogue program - Google Patents

Description

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

In recent years, dialogue systems such as task-oriented dialogue systems and question-and-answer chatbots have been used for FAQ (Frequently Asked Questions) systems, web page searches, and the like. For example, in order to develop a natural topic, prepare multiple scenarios, select a scenario according to the contents of the user utterance uttered by the user, etc., and execute the utterance according to the selected scenario. , systems that interact with users are known.

JP-A-2003-44088 Japanese Patent Application Laid-Open No. 2003-323388

However, in the above technology, since utterances are executed according to a scenario assumed in advance, if the user utters something that deviates from the scenario, the conversation may end due to an abrupt topic change, etc., resulting in system flexibility. is scarce.

It is also conceivable to prepare in advance various contents for utterance in association with scenarios, but the amount of contents to be prepared is enormous, which increases costs and is not practical. In addition, if only the utterances according to the scenario are used, the dialogue with the user becomes monotonous, and the user's satisfaction may decrease and the dialogue may end prematurely.

An object of one aspect is to provide a dialogue device, a dialogue method, and a dialogue program capable of enhancing the continuity of natural dialogue.

In the first proposal, the dialogue device has a storage unit that stores content in which utterance contents to be responded to in response to the user's utterance are defined. The dialogue device has a determination unit that determines depth, which is the degree of detail of the utterance content, based on the user's utterance. The dialogue device has a selection unit that selects utterance target content from the content stored in the storage unit according to the determined depth. The dialogue device has an utterance execution unit that utters an utterance to the user based on the selected utterance target content.

On one side, the continuity of natural dialogue can be enhanced.

FIG. 1 is a diagram illustrating an overall configuration example of a dialogue system according to a first embodiment; FIG. 2 is a functional block diagram of the functional configuration of the interactive device according to the first embodiment; FIG. 3 is a diagram illustrating an example of the data structure of a content DB according to the first embodiment; FIG. 4 is a diagram for explaining the hierarchical structure of content according to the first embodiment. FIG. 5 is a diagram for explaining an increase or decrease in utterance candidates depending on depth. FIG. 6 is a flowchart showing the flow of processing. FIG. 7 is a diagram for explaining an example of dialogue when using the general technique. FIG. 8 is a diagram for explaining an example of dialogue when using the first embodiment. FIG. 9 is a diagram illustrating topic search at low depth. FIG. 10 is a diagram for explaining content selection using keywords. FIG. 11 is a flowchart illustrating the flow of processing according to the second embodiment; FIG. 12 is a diagram illustrating a hardware configuration example.

Hereinafter, embodiments of the interactive device, the interactive method, and the interactive program disclosed in the present application will be described in detail based on the drawings. In addition, this invention is not limited by this Example. Moreover, each embodiment can be appropriately combined within a range without contradiction.

[overall structure]
FIG. 1 is a diagram illustrating an overall configuration example of a dialogue system according to a first embodiment; As shown in FIG. 1, in this interactive system, a user terminal 1 and an interactive device 10 are connected via a network N so as to be communicable. Examples of dialogue include questions from end users such as mail-order sales, net sales, and net games, questions from managers such as system failures, and searches for websites desired by users such as travel destinations.

It should be noted that the network N can employ various networks such as the Internet and dedicated lines, regardless of whether they are wired or wireless. Also, although an example in which one user terminal 1 is used is shown, the present invention is not limited to this, and a plurality of user terminals 1 and the interactive device 10 can simultaneously interact with each other.

The user terminal 1 is a computer device such as a smart phone, a mobile terminal, or a personal computer. The user terminal 1 uses a web browser, a dedicated application, or the like to access the interactive device 10 and interact with the interactive device 10 . In addition, the user terminal 1 executes dialogue by voice input or character input using a keyboard or the like.

The dialogue device 10 is a computer device that establishes Web communication or the like with the user terminal 1 and executes dialogue with the user terminal 1 . This dialogue device 10 realizes a dialogue system such as a task-oriented dialogue system or a question-and-answer type chatbot, and provides topics to the user from the system side.

In such a dialogue system, the dialogue device 10 hierarchizes and holds contents, which are utterance candidates, according to the degree of detail in the topic. Then, based on the topic of the dialogue and the depth of the content used in the dialogue, the dialogue device 10 controls the change and deepening of the topic in shallow hierarchies, and strengthens the limit on the degree of matching with the user's utterance in deep hierarchies. As a result, the dialogue device 10 can develop topics without abruptness, thereby enhancing the continuity of natural dialogue.

[Function configuration]
FIG. 2 is a functional block diagram showing the functional configuration of the interactive device 10 according to the first embodiment. As shown in FIG. 2, the interactive device 10 has a communication section 11, a storage section 12, and a control section 20. FIG.

The communication unit 11 is a processing unit that controls communication with the user terminal 1, and is, for example, a communication interface. For example, the communication unit 11 establishes Web communication or communication with the user terminal 1 using a chatbot, and transmits and receives data related to dialogue.

The storage unit 12 is an example of a storage device that stores programs and data, such as a memory or a hard disk. This storage unit 12 stores a content DB 13 and a dialogue history DB 14 .

The content DB 13 is a database that stores content to be spoken. More specifically, the content DB 13 stores contents in which speech content and the like are designated, and are hierarchically arranged according to the depth indicating the detail degree of the content, which is the prescribed degree of detail of the speech content. FIG. 3 is a diagram illustrating an example of the data structure of the content DB 13 according to the first embodiment; As shown in FIG. 3, the content DB 13 stores content in a hierarchical structure of level 1 content, level 2 content, and level 3 content.

Level 1 content includes a plurality of contents for which "utterance contents, utterance conditions, topic/depth control" are set. In the “utterance content”, the content that the interactive device 10 speaks to the user terminal 1 is set. In the "utterance condition", conditions for selecting content are set, such as levels and keywords. For example, content set at level 1 is preferentially selected when no topic is set. It also indicates that content with the keyword “cooking” is preferentially selected when the latest user utterance includes “cooking”. In "Topic/Depth Control", a control method for topic and depth after content selection is set. For example, for the content set to "If you respond with X1, T1 topic level 2" is set, the topic is set to "T1" when "X1" is included in the user's utterance to the latest content, and the utterance This indicates that the next utterance target is level 2 content for which topic T1 is set as a condition.

The level 2 content has a plurality of contents for which "utterance contents, utterance conditions" are set. In the “utterance content”, the content that the interactive device 10 speaks to the user terminal 1 is set. In the "utterance condition", a condition for selecting content is set, for example, a combination of level and topic, a keyword, and the like are set. For example, content of level 2 or higher and for which topic T is set is preferentially selected when content of level 2 or higher has already been spoken and T is set to the topic. It also indicates that the content with the keyword "curry" is preferentially selected when "curry" is included in the latest user utterance.

The level 3 content has a plurality of contents for which "utterance contents, utterance conditions" are set. In the “utterance content”, the content that the interactive device 10 speaks to the user terminal 1 is set. In the "utterance condition", a condition for selecting content is set. For example, a combination of a level, a topic, and user's utterance contents, a keyword, and the like are set. For example, content of level 2 or higher, topic T, and user utterance Y are set, the content of level 2 or higher has been uttered, T is set to the topic, and "Y" is set to the immediately preceding user utterance. If it is included, it indicates that it is preferentially selected. It also indicates that the content for which the keyword ``the etymology of curry'' is set is preferentially selected when the latest user utterance includes ``the etymology of curry''.

As described above, the content stored in the content DB 13 is divided into levels to form a hierarchical structure. FIG. 4 is a diagram for explaining the hierarchical structure of content according to the first embodiment. As shown in FIG. 4, the hierarchical structure of the content ranges from level 1, which is the shallowest, to level 3, which indicates the degree of detail of the content. Level 1 includes content that can be spoken at any time, Level 2 includes content whose utterance is limited by topic, and Level 3 includes content that responds to user utterances.

A specific example will be described in which the details of the content become more detailed depending on the depth. For example, when a dialogue is started and the topic is not set, the level 1 content of the utterance "I like dogs, but are you a dog person or a cat person?" When the user utters "dog" as , level 2 content with "dog" set as the topic is uttered. For example, as level 2 content, the utterance content "In Japan, toy poodles that are easy to keep in apartments are popular, aren't they?"

After that, when the topic is "dog" and the user makes an utterance including "toy poodle" such as "I like toy poodles" in response to the above utterance, as a response to the user utterance, Level 3 content is spoken. For example, as level 3 content, the utterance content "It is said that poodles with two or more coat colors are popular overseas and are called party colors."

Returning to FIG. 2 , the dialogue history DB 14 is a database that stores the history of dialogue with the user terminal 1 . For example, the dialogue history DB 14 stores the contents of user utterances, the contents of utterances by the dialogue device 10, and the contents of set topics in chronological order. Information stored in the dialogue history DB 14 is saved while the dialogue with the user terminal 1 continues, and is deleted when the dialogue ends.

The control unit 20 is a processing unit that controls the entire interactive device 10, such as a processor. The control section 20 has a content management section 21 , a dialog input section 22 , a dialog control section 23 and a dialog output section 24 . Note that the content management unit 21, the dialogue input unit 22, the dialogue control unit 23, and the dialogue output unit 24 are an example of an electronic circuit such as a processor or an example of a process executed by the processor.

The content management unit 21 is a processing unit that manages content stored in the content DB 13 . Specifically, the content management unit 21 automatically generates hierarchically structured content and stores it in the content DB 13 , or acquires hierarchically structured content generated by an administrator and stores it in the content DB 13 .

For example, the content management unit 21 can acquire content from a web page, set transitions from the top page to a hierarchy, and generate content with a hierarchical structure. Taking a travel site as an example, the content on the top page, "hot spring special", is level 1 content, and the content that transitions from the top page to the next is "simple hot spring," "acidic spring," and "open-air bath with a view." is level 2 content. In addition, the pages of each inn, such as the content "○○ Inn", which transitions from the content "simple hot spring", "acidic spring", and "open-air bath with a view", are level 3 content.

Here, a method for setting the topic and depth by the content management unit 21 will be described. For example, the content management unit 21 creates content that describes another condition such as a user utterance in addition to a depth condition lower than the depth X in the utterance condition of the content of depth X. Control information can be described in the content of the depth. In addition, when the interests of the user match, the content management unit 21 assigns a keyword to each content in order to make it easier to deepen the topic. can also be written as

The dialogue input unit 22 is a processing unit that acquires user speech uttered from the user terminal 1 and inputs it to the dialogue control unit 23 . For example, the dialog input unit 22 acquires a user utterance such as “Tell me something”, inputs it to the dialog control unit 23, and stores it in the dialog history DB 14. FIG. The dialog input unit 22 can also perform morphological analysis and the like on the contents of user utterances, and input information such as extracted words to the dialog control unit 23 as analysis results.

The dialog control unit 23 has a topic management unit 23a, a candidate extraction unit 23b, and a selection unit 23c, and is a control unit that identifies relevant content according to user utterances and executes the utterances to the user.

The topic management unit 23a is a processing unit that manages topics of dialogue based on the state of dialogue with the user. Specifically, the topic management unit 23a manages the state of topics and the level of depth. In addition, the topic management unit 23a acquires and holds information on the content spoken from the dialogue output unit 24. FIG. Then, the topic management unit 23a determines the topic and level based on the immediately preceding content, the depth control information, and the content of the user's utterance.

For example, when level 1 content A is uttered, the topic management unit 23a holds the topic and depth control set for the content "If X1 is answered, T1 topic level 2". After that, when the user's utterance includes "X1", the topic management unit 23a sets the topic to "T1" and sets the dialogue level, which is the degree of detail of the dialogue with the user, to level 2. On the other hand, if "X1" is not included in the user's utterance, the topic management unit 23a leaves the topic unset and leaves the conversation level at level 1.

Also, the topic management unit 23a maintains the dialogue level at level 2 when a user response similar to the depth control set for content A at level 1 is detected. Content of level 1 can be uttered even when the content is of level 2 or higher, and if the condition for determining the depth is met, a new topic of level 2 is set. Further, when the speech time with the user has passed for a predetermined period of time or when the user's response is low, the topic is not set.

The candidate extraction unit 23b is a processing unit that extracts contents of utterance candidates based on topics and utterance conditions. Specifically, when the level of the content that has already been spoken is low (low level), the candidate extraction unit 23b extracts the content of the speech candidate so as to search for a topic with more options. In addition, when the level of the uttered content is deep (high level), the candidate extraction unit 23b extracts the utterance candidate content so as to follow the content of the user's utterance. Then, the candidate extraction unit 23b outputs the extracted result to the selection unit 23c.

FIG. 5 is a diagram for explaining an increase or decrease in utterance candidates depending on depth. As shown in FIG. 5, the candidate extracting unit 23b extracts level 1 content as an utterance candidate when there is no topic. Further, when the topic is "X", the candidate extraction unit 23b extracts the content of level 1 and the content of level 2 to which the topic "X" is set as utterance candidates. In this case, when topic X is set, a dialogue related to X is taking place, and content of level 2 contains keywords related to X, so it is more likely to be selected than level 1 candidates. Further, when the topic is "X" and the user utterance is "aa", the candidate extraction unit 23b adds the level 1 content and the level 2 content set with the topic "X" to the topic "X". In addition, the content of level 3 in which the user's utterance is set to "aa" is extracted as an utterance candidate. Thus, as the depth increases, more detailed content is likely to be selected.

More specifically, when the topic is not set or when the dialogue level is 1, the candidate extraction unit 23b extracts level 1 content as an utterance candidate. Further, the candidate extracting unit 23b, after the content A of level 1 is uttered, when the user utterance corresponding to the topic and the depth control set for the content "If you respond with X1, the level 2 of the T1 topic" is executed , "T1" is set as the topic by the topic management unit 23a, so in addition to the content of level 1, the content of level 2 whose topic is "T1" is extracted as an utterance candidate according to the depth control.

Further, when the topic is “T” and the dialogue level is 2 or higher, and the latest user utterance includes “Y”, the candidate extraction unit 23b selects “topic is T and Contents having an utterance condition that the dialogue level is 2 or more and Y is included in the user's utterance are extracted as utterance candidates. Further, when the user utters a keyword that matches the utterance condition, the candidate extraction unit 23b limits output candidates to content that matches the keyword.

Further, when the user utters a negative utterance such as "not interested", the candidate extraction unit 23b lowers the level while maintaining the topic, and selects utterance candidate content. Further, when the topic management unit 23a changes the topic to unset, the candidate extraction unit 23b extracts content of level 1 as an utterance candidate. That is, as the topic is changed to unset, the utterance level is lowered and the topic is reexamined.

The selection unit 23c is a processing unit that selects utterance target content from the utterance candidate content extracted by the candidate extraction unit 23b. Specifically, the selection unit 23c selects utterance candidates from the contents of the utterance candidates using a method such as keyword matching or random, and outputs the selected utterance candidates to the dialogue output unit 24 .

For example, when the topic is not set, the selection unit 23c can randomly select content of level 1, which is an utterance candidate. For example, when the number of consecutive utterances of level 1 content is less than a threshold, the selection unit 23c selects content that has a common category with output content so as not to cause a large topic change. More specifically, when the content of "baseball" has already been output, the selection unit 23c selects the content of "soccer", which is the same sport.

Further, when the dialogue level is 2, the selection unit 23c can preferentially select the content of level 2 among the content of the utterance candidates. For example, when the topic is "cooking", the selecting unit 23c selects content whose utterance content is a cooking name such as "curry" or "sushi", which is a subordinate concept of "cooking" in more detail.

Moreover, when the content of level 3 is an utterance candidate, the selection unit 23c can preferentially select the content that follows the user's utterance. For example, when the topic is "cooking" and the history of user utterances includes "curry", the selection unit 23c selects "how to make curry" and "countries where curry can be eaten" so as not to deviate from the topic of "curry". Select content that sets detailed contents about "curry" such as ".

Returning to FIG. 2, the dialogue output unit 24 is a processing unit that outputs the content selected by the selection unit 23c. Further, the dialogue output unit 24 outputs information about content output (spoken) to the user terminal to the topic management unit 23a and stores the information in the dialogue history DB 14. FIG.

[Process flow]
FIG. 6 is a flowchart showing the flow of processing. As shown in FIG. 6, when the dialog input unit 22 receives a user reaction (S101: Yes), the dialog control unit 23 updates the topic according to the user reaction, and sets or deletes the topic (S102). .

Subsequently, the dialogue control unit 23 determines the depth (S103), extracts utterance candidate content according to the depth (S104), and selects utterance target content from among the utterance candidates based on criteria such as randomness (S105). ). After that, the dialogue output unit 24 outputs the speech content set in the selected speech target content to the user terminal 1 (S106).

[Comparative example]
Next, a comparative example between the general technique and Example 1 will be described. FIG. 7 is a diagram for explaining an example of dialogue when using the general technique, and FIG. 8 is a diagram for explaining an example of dialogue when using the first embodiment.

As shown in (1) of FIG. 7, when a user requests a topic such as "Tell me something," the general system using the general technology responds to a scenario, recent trends, etc., and responds with a "night pool." ”. Subsequently, as shown in (2) of FIG. 7, the general system repeats a specific utterance indicating a backtracking such as uttered by the user when the user's interest is low or when the scenario deviates. Even in the case where the user has been asked to do so, he or she simply backtracks by uttering a pre-designated phrase such as "that's right".

After that, as shown in (3) of FIG. 7, since the user uttered a content such as "What is that?" that the user utters when he/she is in a bad mood, the general system suddenly perform the conversion. As a result, as shown in (4) of FIG. 7, the user's satisfaction is lowered, and the dialogue is forcibly terminated.

On the other hand, as shown in (1) of FIG. 8, when the user requests a topic such as "Tell me something", the interactive device 10 according to the first embodiment displays a topic such as "video game" or "cooking". Speak level 1 content that includes the topic of Subsequently, as shown in (2) of FIG. 8 , when a user utterance including “cooking” is input, the interactive device 10 sets “cooking” to the topic, and “cooking” is associated with the topic “cooking”. Speak level 2 content for inquiring "favorite dish name". In response to this utterance, the interactive device 10 specifies that the favorite dish name is "curry".

After that, as shown in (3) of FIG. 8, since "curry" which is a more specific representation of "cooking" was uttered, the interactive device 10 selects "curry Speak level 3 content that explains the etymology of As a result, as shown in (4) of FIG. 8, it is possible to suppress inadvertent topic changes and inorganic dialogues according to the scenario, thereby improving the user's satisfaction.

[effect]
As described above, the interactive device 10 determines the current topic and depth from the output of content and the history of user responses, extracts output candidates from the current topic and depth, and selects and outputs content from the output candidates. . In this way, the dialogue device 10 holds utterance candidates in a hierarchical manner according to the degree of detail in the topic, and each content has information for determining utterance conditions and depth. Therefore, the dialogue apparatus 10 can increase matching conditions with topics and user utterances for utterance candidates in a deep hierarchy, and control change and deepening of topics based on the topics.

As a result, the dialogue device 10 can realize dialogues that are friendly to the user by less abrupt topic changes and deepening of topics, and can enhance the continuity of natural dialogues. In addition, the dialogue apparatus 10 searches for topics at a shallow level, and deepens the dialogue at a deep level, thereby improving the user's satisfaction. In addition, the interactive device 10 can reduce the cost of creating content in which the contents are gradually detailed, compared to a scenario format that requires a flow.

By the way, in the first embodiment, an example has been described in which utterance restriction and utterance control are realized by setting the utterance condition on the content side, but the present invention is not limited to this. For example, the information set on the content side can be reduced to reduce the burden on the content creator, and the dialog device 10 can take the lead in executing content selection and utterance control using the topic search status and keyword utterance status. can.

FIG. 9 is a diagram for explaining topic search in low depth, and FIG. 10 is a diagram for explaining content selection by keyword. As shown in FIG. 9, upon receiving the user's utterance, the dialogue control unit 23 of the dialogue device 10 utters content highly relevant to the user's utterance or randomly selected content from the level 1 contents (( 1) see). Note that highly relevant content includes, for example, content belonging to the same genre or topic as the spoken content, content including words included in the spoken content, synonyms of the words, and the like. Also, the content with low relevance is, for example, content belonging to a genre or topic different from that of the spoken content, content that does not include words included in the spoken content, synonyms of the words, and the like.

Subsequently, since the user's utterance, which is a response to the uttered content, does not include a word corresponding to the predetermined topic, the dialogue control unit 23 selects and utters the content of level 1 again (FIG. 9). (2)). At this time, the interaction control unit 23 can also select content at random, or select content that has a low degree of similarity to content that has already been spoken. Note that the degree of similarity can be determined based on whether or not it is equal to or greater than a threshold using the degree of matching between words.

Similarly, since the user's utterance, which is a response to the uttered content, does not contain a word corresponding to the predetermined topic, the dialogue control unit 23 selects and utters the content of level 1 again (FIG. 9). (3)). In this way, the dialogue control unit 23 searches for topics that the user is interested in.

After that, when the dialog control unit 23 detects that the user's utterance, which is a response to the uttered content, includes a word Y corresponding to a predetermined topic, it sets Y as the topic of the dialog with the user. (See (4) in FIG. 9). Here, in addition to the level 1 content, the dialogue control unit 23 sets content including the word Y corresponding to the topic Y among the level 2 content as an utterance candidate. Note that the level 2 content is preferentially selected, and the level 1 content is selected when there is no corresponding content or when the utterance target level 2 content has already been uttered. Level 2 may also be categorized by topic.

Then, the dialog control unit 23 selects the level 2 content that includes the word Y or the content that corresponds to the topic Y, and speaks to the user (see (5) in FIG. 9). In this way, the dialogue device 10 searches for a topic in which the user is interested, and when the topic in which the user is interested is searched, utters more detailed content about the topic, thereby reflecting the content of the dialogue with the user. can be enriched.

Further, as shown in FIG. 10, upon receiving the user's speech, the interaction control unit 23 randomly selects content of level 1 and speaks (see (1) in FIG. 10). As a response, when the user utters an utterance containing the keyword "KY1", the interaction control unit 23 detects a content group of level 3 in which the keyword "KY1" is set, and randomly selects content from the content group. (see (2) in FIG. 10). In this way, the dialogue control unit 23 preferentially uses keywords to select content regardless of the level. Further, the dialogue control unit 23 sets the topic set in the spoken level 3 content or the word included in the level 3 content as the current topic.

Next, after the level 3 content is uttered, the dialogue control unit 23 determines that the user is showing interest while the user does not make a mere backtrack or an utterance indicating the end of the dialogue, and utters the keyword "KY1". Select the content of level 3 that the user has or the content of level 2 that matches the topic, and continue speaking (see (3) in FIG. 10). At this time, the interaction control unit 23 can select not only the keyword included in the user's utterance but also the content similar to the uttered content or the content of level 2 or 3 belonging to the same category.

After that, when a user utterance such as "I'm not interested" is input to the uttered level 3 content, the dialogue control unit 23 determines that the user is no longer interested, cancels the setting of the topic, The content of level 1 is randomly selected and spoken (see (4) in FIG. 10).

In this way, even when the user is speaking the content of level 3, if the user loses interest, the dialogue device 10 speaks the content of level 1 again, thereby increasing the interest of the user. Since the topic is searched again, it is possible to suppress the decrease in user's satisfaction.

FIG. 11 is a flowchart illustrating the flow of processing according to the second embodiment; As shown in FIG. 11, upon receiving a user's utterance (S201: Yes), the dialogue control unit 23 of the dialogue device 10 determines whether or not it matches a negative utterance specified in advance (S202). For example, the dialog control unit 23 can preliminarily designate utterances such as "I'm not interested", "Is that so?", "Huh?"

Subsequently, if the user's utterance is not a negative utterance (S202: No), the dialogue control unit 23 determines whether or not it matches the utterance condition of the content of level 3 (S203). For example, the dialogue control unit 23 determines whether or not the keyword set for the level 3 content is included in the user's utterance. As another example, the dialog control unit 23 determines whether or not the user's utterance includes a keyword (word) that is a lower concept than the keyword (word) included in the level 2 content that was uttered immediately before. . Further, the dialogue control unit 23 determines whether or not the content of level 2 is continuously uttered a plurality of times.

Then, when matching the utterance condition of the content of level 3 (S203: Yes), the dialogue control unit 23 selects content that makes the content of level 3 more interesting (S204). For example, the dialogue control unit 23 randomly selects content, content similar to the content uttered immediately before, content belonging to the same category as the content uttered immediately before, and keywords contained in the content uttered immediately before. Select content that contains similar keywords or subordinate keywords.

On the other hand, if the speech condition does not match the utterance condition of the content of level 3 (S203: No), the dialogue control unit 23 determines whether or not the topic has been decided (S205).

Then, when the topic has been decided (S205: Yes), the dialogue control unit 23 selects content that makes the user more interested from the level 1 content and the corresponding level 2 content (S206). For example, the dialog control unit 23 can be selected based on the same criteria as in S204.

On the other hand, if the topic has not been finalized (S205: No), the dialogue control unit 23 selects unuttered content from the content of level 1 (S207). Note that S207 is also executed when it is determined in S202 that the user's utterance is a negative utterance (S202: Yes).

After that, the dialogue control unit 23 utters the utterance content set in the content selected in any of S204, S206, and S207 to the user terminal (S208).

As described above, the dialog apparatus 10 can take the lead in executing content selection and utterance control using the topic search state and keyword utterance state. The burden can be reduced.

Although the embodiments of the present invention have been described so far, the present invention may be implemented in various different forms other than the embodiments described above.

[level]
For example, in the above embodiment, an example of a three-stage hierarchical structure from level 1 to level 3 has been described, but the present invention is not limited to this, and a hierarchical structure of two stages or four or more stages can also be used. Also, in the embodiment, for the sake of explanation, an example in which a level is explicitly set for each content has been described, but the present invention is not limited to this. For example, by setting utterance contents step by step, such as "cooking", "curry", and "how to make curry", it is possible to configure the content so that it results in a hierarchical structure.

[content]
The data structure of the content and setting examples described in the above embodiments are only examples, and the settings can be changed arbitrarily. For example, it is possible to set multiple topics or set different utterance conditions for each of the multiple topics. Also, known contents such as dialog blocks used in dialog systems can be employed.

The dialogue device 10 can also set multiple topics during one dialogue. In this case, the dialogue device 10 allows the user to select one of the dialogues, give priority to the latest topic, or give priority to the topic that the user has spoken many times. .

[Depth control]
For example, the dialogue device 10 can control the topic and depth depending on the content of the utterance without depending on the user's response. Specifically, the interactive device 10 reduces the topic depth when the user is not interested. In addition, the dialogue device 10 deletes the topic when the user utters a preset specific word such as "It's okay", "Stop", "Tell me something else", and deletes the content of level 1. You can also choose from

In addition, the dialogue device 10 can also transition to a level one level lower than the current level when the same level of content is spoken for a predetermined time such as three minutes or more. The interactive device 10 can also set the user's emotion as the depth condition of the content. In the case of voice input, the emotion can be the volume of the voice, each frequency indicating the emotion, or the like. For example, if a negative emotion such as anger or sadness is detected, the level is lowered by one step to make the hierarchy shallower, and if a positive emotion such as joy or surprise is detected, the level is raised by one step to make the hierarchy can also be deepened.

[system]
Information including processing procedures, control procedures, specific names, and various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified.

Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. That is, the specific forms of distribution and integration of each device are not limited to those shown in the drawings. That is, all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. For example, the content management section 21 and the dialog control section 23 can be integrated.

Further, each processing function performed by each device may be implemented in whole or in part by a CPU and a program analyzed and executed by the CPU, or implemented as hardware based on wired logic.

[hardware]
FIG. 12 is a diagram illustrating a hardware configuration example. As shown in FIG. 12, the interactive device 10 has a communication device 10a, a HDD (Hard Disk Drive) 10b, a memory 10c, and a processor 10d. 12 are interconnected by a bus or the like.

The communication device 10a is a network interface card or the like, and communicates with other servers. The HDD 10b stores programs and DBs for operating the functions shown in FIG.

The processor 10d reads from the HDD 10b or the like a program that executes the same processing as each processing unit shown in FIG. 2 and develops it in the memory 10c, thereby operating the process of executing each function described with reference to FIG. 2 and the like. That is, this process executes the same function as each processing unit of the interactive device 10. FIG. Specifically, the processor 10d reads a program having functions similar to those of the content management unit 21, the dialogue input unit 22, the dialogue control unit 23, the dialogue output unit 24, and the like, from the HDD 10b and the like. Then, the processor 10d executes the same processes as those of the content management unit 21, the dialogue input unit 22, the dialogue control unit 23, the dialogue output unit 24, and the like.

Thus, the interactive device 10 operates as an information processing device that executes an interactive method by reading and executing a program. Further, the interactive device 10 can read the program from the recording medium by the medium reading device and execute the read program to realize the same functions as the above embodiment. It should be noted that the programs referred to in other embodiments are not limited to being executed by the interactive device 10. FIG. For example, the present invention can be applied in the same way when another computer or server executes the program, or when they cooperate to execute the program.

10 dialogue device 11 communication unit 12 storage unit 13 content DB
14 Dialogue history DB
20 control unit 21 content management unit 22 dialogue input unit 23 dialogue control unit 23a topic management unit 23b candidate extraction unit 23c selection unit 24 dialogue output unit

Claims (10)

A content in which utterance content to be responded to by a user is defined, wherein the topic of dialogue and the utterance content to respond to the dialogue are limited for each hierarchical level as the depth of the level increases. a storage unit that stores content ;
a decision unit that decides, based on the user's utterance, the occurrence status of the topic of the dialogue and the status of the level used in the response of the dialogue ;
In the interaction with the user, if the topic has not been set, the shallow level content used for setting the topic is selected, and if the topic has been set, the topic has been set. a selection unit that selects content at a deep level according to the depth and appearance of keywords in the dialogue ;
and an utterance execution unit that utters to the user using the utterance content set in the selected content . 2. The storage unit according to claim 1, wherein the deeper the hierarchies of the depths are, the more closely related to the utterance content of the user is stored as the utterance target content. interactive device. The storage unit stores the content for each depth layered according to the defined degree of detail of the utterance content,
2. The dialogue apparatus according to claim 1, wherein the selection unit selects content to be spoken from among a plurality of contents corresponding to the hierarchy of depth corresponding to the determined level . The storage unit stores the content in which the utterance content and utterance conditions are set for each depth hierarchized according to the defined degree of detail of the utterance content,
The selection unit selects content satisfying the utterance condition as content to be uttered based on the user's response to the uttered content and the depth hierarchy to which the uttered content belongs. 2. A dialogue device according to claim 1, characterized in that it selects. The depth is hierarchized by levels indicating the degree of detail of interaction with the user,
In the utterance condition, the higher the level, the more detailed conditions are specified that make it difficult to utter,
The determining unit determines the level based on the user's response to the spoken content and the depth hierarchy to which the spoken content belongs,
5. The dialogue apparatus according to claim 4, wherein the selection unit selects content satisfying the utterance condition from among contents corresponding to the determined level as the utterance target content. The utterance condition of the contents of each level includes keywords,
wherein, when the keyword is included in the response content of the user, the selection unit selects the content to which the keyword is set as the content to be uttered regardless of the specified level. Item 6. An interactive device according to item 5. When the content of a response from the user includes a negative word, or when the time of interaction with the user exceeds a predetermined time, the selection unit selects content belonging to a level shallower than a specified level, 6. The interactive device according to claim 5, wherein the content to be spoken is selected. 2. The dialogue according to claim 1, wherein the storage unit stores, as the utterance target content, the shallower the layered depth, the lower the relevance to the uttered content. Device. the computer
determining, based on user utterances, the occurrence of a topic of dialogue and the level of context being used in the response of said dialogue ;
The content defines utterance content to be responded to according to the user's utterance, wherein the topic of dialogue and the utterance content to respond to the dialogue are limited for each hierarchical level as the depth of the level increases. By referring to a storage unit that stores each content , if the topic has not been set in the dialogue with the user, the shallow level content used for setting the topic is selected, and the topic is set. if it has already been done, selecting content at a deeper level in accordance with the depth at which the topic is set and the occurrence of keywords in the dialogue ;
A dialogue method characterized by executing a process of uttering to the user using the utterance content set in the selected content . to the computer,
determining, based on user utterances, the occurrence of a topic of dialogue and the level of context being used in the response of said dialogue ;
The content defines utterance content to be responded to according to the user's utterance, wherein the topic of dialogue and the utterance content to respond to the dialogue are limited for each hierarchical level as the depth of the level increases. By referring to a storage unit that stores each content , if the topic has not been set in the dialogue with the user, the shallow level content used for setting the topic is selected, and the topic is set. if it has already been done, selecting content at a deeper level in accordance with the depth at which the topic is set and the occurrence of keywords in the dialogue ;
A dialog program characterized by executing a process of speaking to the user using the speech content set in the selected content .

Images