JP2003058187A - Interaction system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an interaction system with which a degree of freedom, flexibility and easiness in interaction design are improved. SOLUTION: In an interaction system 10 provided with a voice recognizing processing part 11, an interaction engine 12, an interaction database 15 and a voice synthesizing part 16 at least, an interaction transition control means 21 is provided for specifying the transition destination of interaction corresponding to the filling degree of interaction data for successively filling a series of slots forming frames comprising the interaction.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dialogue system, for example, a dialogue system as a guidance function unit in a navigation device which is one of in-vehicle information devices and an automatic customer service device in a fast food restaurant.

[0002]

2. Description of the Related Art In order to fulfill the function of the above guidance,
The dialogue system includes a voice recognition processing unit, a dialogue engine, a voice synthesis unit, a dialogue database, and the like, and dialogue data processing is performed in the system.

In this case, the voice synthesizing unit makes an inquiry to the user as voice guidance, and when the user utters in response to this, the voice recognition processing unit analyzes the meaning of the utterance. Then, the dialogue engine generates the next voice guidance corresponding to the user's utterance according to the application (for example, the above-mentioned navigation or automatic customer service) according to the analysis result, and again asks the user a new question through the voice synthesis unit. To do. After that, the flow of the dialogue is completed by repeating the same process. The flow of this dialogue is performed by the internal data processing of the system.

Although there are various modes of executing the above internal data processing, in the present invention, so-called "frame processing" is performed.
We describe a dialogue system based on the form of execution by. The frame in this "frame processing" is
It is a series of slots, and one example of navigation is "How many prefectures do you go to?" → Slot <2> = "Hyogo prefecture" → "Which city in Hyogo prefecture do you go to" → Slot <4> = “Kobe City” → “Tell me what kind of facility” → Slot <6> = A series of slots <2><4><6> such as “Port Island” is one frame. Assembling is called frame processing.

[0005] Or, to give an example of a fast food restaurant, "what do you want for food?" → slot <2
> = “Hamburger” → slot <3> = “2” →
Frame processing is the assembly of a series of slots <2><3> such as "Did you eat here?" ...

Regarding the frame processing, conventionally, the following processing is included.

1) A predetermined series of slots forming a frame of a certain dialogue is filled with corresponding utterance data (the above-mentioned "Hyogo prefecture", etc.), thereby transitioning to the next dialogue.

2) After all of the series of slots are filled,
A fixed transition is made to one predetermined transition destination dialogue.

3) When the constraint of 2) above is released, the program is described in the dialogue transition data, that is, the automaton.

[0010]

1) Due to the above 1), it is impossible to transit to the next dialogue unless all the series of slots are filled with the respective speech data.
In other words, there is a problem that the degree of freedom in dialogue design is extremely small.

2) Due to the above 2), when all of the series of slots are filled with the respective speech data, only one dialogue can be transitioned to next. In other words, there is a problem that the flexibility of dialogue design is lacking.

3) Due to the above 3), since the description of the automaton is mixed, it is not possible to uniformly process data in all dialogues by frame processing. That is, there is a problem that the dialogue design becomes complicated.

4) Further, since the description of the above automaton is mixed, there is a problem that the data size of each utterance data used for the above internal data processing becomes large.

Therefore, in view of the above problems, the present invention provides a dialog system which can realize all dialogs by frame processing and can simultaneously satisfy the degree of freedom, flexibility and simplicity of dialog design. It is intended.

[0015]

FIG. 1 is a diagram showing an example of the configuration of a dialogue system to which the present invention is applied.

In the figure, the dialogue system 10 also includes a voice recognition processing section 11 which analyzes the user's utterance input from the user through the microphone MC and recognizes its meaning while referring to the recognition dictionary 14, and the recognition result. In addition, according to the application 13 (navigation and automatic customer service described above), the dialogue engine 12 that generates a flow of dialogue while referring to the dialogue database (dialogue DB) 15, the generated dialogue is synthesized as voice and provided to the user from the speaker SP. Display the screen related to the dialogue with the voice synthesizer 16
GUI (Graphical User)
Interface) control unit 17.

Of these components 11-18, of particular relevance to the present invention is the dialogue engine 12 and the dialogue database 15 which cooperates with the dialogue engine 12.

FIG. 2 is a diagram showing a main part of the present invention.

Particularly, the portions of the dialogue engine 12 and the dialogue database 15 in FIG. 1 are taken out and specifically shown.

The component of particular interest in FIG. 2 is the dialogue control means 21. That is, it is a dialogue transition control means for specifying the transition destination of the dialogue according to the filling degree of the dialogue data that sequentially fills a series of slots that form the frame that constitutes the dialogue.

Further noteworthy components are the data area 22 and the frame reference table 23. Specifically, first, the dialogue data corresponding to each of a series of slots forming a frame constituting a dialogue is filled in the frame data area 22 for filling the corresponding slot. A frame reference table 23, which defines transition destinations of dialogues corresponding to each of a plurality of frame patterns determined by the filling degree of the given dialogue data, is constructed in the dialogue database 15 in advance.

Then, using the frame data area 22 and the frame reference table 23, the dialogue transition control means 2 for specifying the dialogue transition destination according to the filling degree of the dialogue data.
1 is formed in the dialogue engine 12.

With this, since the transition destination of the dialogue can be changed according to the filling degree of the slots, the degree of freedom of the dialogue design can be increased and the dialogue design can be made flexible.

[0024]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 3 is a diagram showing a detailed example of the frame data area 22 of FIG.

In the figure, the frame data area 22 is
For the sake of clarity, multiple blocks are stacked and drawn in multiple stages.

A frame forming one dialogue is specified by a dialogue ID and a dialogue state ID.

The dialogue ID is for displaying what kind of dialogue the dialogue system is having now. For example,
This is an ID for representing a dialogue for asking the user about the "destination" or a dialogue for asking the user for "ordering food".

The dialog state ID is used by the dialog system now.
It is for displaying the number of states in one dialogue flow. For example, in the case of a dialogue for asking the user about the "destination", what is the state (state) in the flow of "what prefecture" (first), "what city" (second) ... This is an ID for representing.

A set of the above dialogue ID and dialogue state ID is
When sent from the dialogue engine 12 to the application 13,
The application 13 controls the voice synthesis unit 16 so as to provide the user with guidance corresponding to the set.

FIG. 4 is a diagram showing an example of the dialogue ID and the dialogue state ID.

In the figure, S-0 is the above-mentioned dialogue ID, and the numbers (1, 2, ...) Attached thereto are the above-mentioned dialogue state ID. Note that S on the right side of the figure represents utterance on the system side, and U represents utterance on the user side.

Returning to FIG. 3, the above-mentioned dialogue ID and dialogue state ID are shown in the top row (first row) of the figure.
The number of slots M is described in the second row. It is the number of the already described slot <1>, slot <2>, slot <3>, .... If M = 6, one frame consists of a series of 6 slots.

The third and subsequent stages in FIG. 3 correspond to each slot,
It is composed of M slots 1, 2, ... M.

The slot value attribute 1A in the third row is "prefecture name" in navigation and "food" in automatic customer service.
Is equivalent to a dialogue that asks. Then, "Hyogo Prefecture" and "Hamburger" that answer the answer become "Value 1" shown in the third row.

The slot value attribute 1B associated with the slot value attribute 1A is not always necessary data, but the "prefecture name" which is the slot value attribute 1A is, for example, "destination setting" or " When the application 13 needs to further know whether it is "relay point setting", the slot value attribute 1B displays that "Hyogo prefecture" is "destination setting" or "relay point setting". be able to.

The same kind of data as in the third row is described in the fourth and subsequent rows in FIG.

FIG. 5 is a diagram showing a detailed example of the frame reference table 23 of FIG.

In this figure, the frame reference table 23
Consists of columns of the first stage, the second stage to the Nth stage, and each column corresponds to each frame pattern.

Taking the uppermost stage (first stage) as an example, the frame pattern 1 has a certain conversation ID, for example S-0, and its conversation state ID is 1, for example (S-0: 1). That is, the dialogue data filled in the frame data area 22 corresponds to S-0: 1.

The frame reference table 23 indicates that if the flow of the dialogue becomes such a frame pattern (S-0: 1), "to next dialogue ID / interaction state ID". Is. The content of this instruction is “transition destination dialogue ID” / “transition destination dialogue state ID” in the table 23. The above frame pattern 1 (= S-0:
In the example of 1), for example, S-0: 2 is described in the "transition destination conversation ID" / "transition destination conversation state ID".

FIG. 6 is a diagram schematically showing how to retrieve the frame pattern of FIG.

However, for simplification, the case where one frame consists of three slots is shown. Therefore, there are eight types of frame patterns that can be adopted.

The above-mentioned frame data area 22 (FIG. 3)
In the case where only the slot <1> is filled, the column of the frame pattern 1 is referred to depending on the filling degree of the slot (conversation data). Similarly, slot <1><
2> only, slot <1><3> only, slot <1>
If all of <2> and <3> are filled, the columns of the frame patterns 2, 3 and 8 are referred to respectively. Then, the dialogue is transitioned according to the transition destination ID / transition destination state ID shown in each column. The above is the first of the present invention
It is a mode of.

FIG. 7 is a flow chart showing the operation of the first mode according to the present invention.

In this figure, step S11: the frame data area 22 shown in FIG.
It is the operation in.

Step S12: A search operation of the frame reference table 23 described with reference to FIGS. 5 and 6.

8A and 8B are transition diagrams showing two simple examples of the interaction model according to the first mode.

FIG. 8A shows a state in which the dialogue state is continuously changed within the same dialogue ID according to the utterance content (according to the frame pattern), and FIG. 8B shows the utterance content according to the utterance content. This shows a transition to a conversation ID different from the conversation IDs used so far.

FIG. 9 is a flowchart showing the operation of the second mode according to the present invention.

The feature of this second aspect is that if the dialogue ID of the ongoing dialogue transits to a dialogue ID different from the dialogue ID before the dialogue data is completely filled in all of the series of slots forming the frame, the progress is made so far. The feature is that the frame processing that has been performed is ended.

According to the conventional frame processing, it is not possible to transit to another conversation ID before each conversation data is completely filled in all the series of slots forming a frame. That is, the frame processing cannot be completed unless the dialog data is completely filled.

However, according to the present invention, according to the frame reference table 23 shown in FIG. 5, it is possible to flexibly move to another dialogue. Therefore, when such a dialogue transition occurs, regardless of whether all slots are filled or not. The frame processing up to that point will be terminated.

As a result, the degree of freedom in dialog design is increased, and flexible dialog design is possible.

Referring now to FIG. 9, step S21
And S22 correspond to step S1 shown in FIG. 7, respectively.
1 and S12.

One of the features of the second aspect is that step S2
3 from one dialogue to another (for example, S
-0 → S-1) It is to judge whether to make a transition.

If the conversation ID has not changed (the same conversation is in progress), the frame processing is continued. That is,
If the determination result of step S23 is Yes, step S
Proceed to 24. This is not so different from the conventional one, but the second
A further feature of this mode is that when one dialogue ID transitions to another dialogue ID according to the frame reference table 23 (the determination result in step S23 is No), the frame processing currently in progress ends. Yes (see the first half of each of steps S26 and S27).

In the latter half of each of steps S26 and S27, when the frame reference table 23 (FIG. 5) is referenced, whether or not the "dialogue end code" is entered in the corresponding table portion (step S25). It depends on

That is, the determination result of step S25 is Yes.
In the case of s (there is a dialogue end code), the dialogue is ended. On the other hand, if the determination result in step S25 is No (no dialogue end code), the dialogue moves to the transition destination dialogue and continues.

FIG. 10 is a transition diagram showing an example of a dialogue model according to the second mode.

The following three cases will be described assuming that three slots related to the dialogue data of “affirmation”, “denial”, and “prefecture name” are prepared.

Case 1) In the state of dialogue state 0 in FIG.
If the user utters “Hyogo prefecture” related to “prefecture name”, the slot related to “prefecture name”, which is one of the slots, is filled. Then, the state moves to the dialogue state 3.

In this dialogue state 3, the frame reference table 23 is referred to and the transition destination dialogue ID described therein is written.
And the transition destination conversation state ID.

In this case, it is possible to make a transition to a different dialogue ID even though the slots related to "affirmation" and "negative" are not filled. At this time, the frame before the transition is discarded.

Case 2) In the state of dialogue state 0 in FIG.
If the user utters "no" related to "negative", the slot related to one of the slots, "negative", is filled. Then, the state moves to the dialogue state 2.

In this conversation state 2, the frame reference table 2
3, the transition is made to the transition destination conversation ID and the transition destination conversation state ID described therein.

In this case, it is possible to make a transition to a different dialogue ID even though the slots related to "affirmation" and "prefecture name" are not filled. At this time, the frame before the transition is discarded.

Case 3) In the state of dialogue state 0 in FIG.
If the user utters "yes" related to "affirmation", the slot related to one of the slots, "affirmation", is filled. Frame reference table 2 in this state
3, the transition destination conversation ID described therein:
Transition to the same ID and transition destination conversation state ID: 1.

In this case, after the application guides the user based on the transition ID and the transition state ID data, the transition destination conversation state: 5 is entered.

If the user utters "Hyogo prefecture" related to "prefecture name" in the state of the dialogue state 5, the slot related to "prefecture name" which is one of the slots is filled.
Then, the state moves to the dialogue state 3.

In this interactive state 3, the frame reference table 2
3, the transition is made to the transition destination conversation ID and the transition destination conversation state ID described therein.

In this case, it is possible to make a transition to a different conversation ID even though the slot related to "negative" is not filled. At this time, the frame before the transition is discarded.

FIG. 11 is a diagram showing a third mode according to the present invention.

As shown in the figure, the frame reference table 23 based on the third mode has a transition destination ID area 25 (see FIG. 5).
The same) and the extension area 26.

That is, in the third mode, each frame pattern in the frame reference table 23 is added to the transition destination ID area 25 consisting of the transition destination conversation ID and the transition destination conversation state ID defined for each frame pattern. It is characterized by having an extension area 26 which describes a transition destination to a specific event defined for each frame pattern. In FIG. 11, the extension area 26 is shown as a pair of extension areas I and II.

The above specific event will now be described with reference to the first and second examples.

According to the first example, the event is an assisted dialogue. The assisted dialogue is such that there is no response from the user even if a predetermined time (for example, 5 seconds) elapses in response to an inquiry from the system side, or a response from the user requests “help”. In the case where there is such a situation, it is a dialogue for re-questioning in a form that is easier for the system to answer.

Therefore, the conversation ID of the above assisted dialogue is described in the extended area I, and the dialogue state ID of the assisted dialogue can be described in the extended area II. The dialogue ID and the dialogue state ID differ depending on the progress of the dialogue, that is, the type of frame pattern.

According to the second example, the event is, for example, an event for shifting to an application, for example, a control code for starting a timer.

Thus, according to the third aspect, it is possible to improve the degree of freedom and flexibility of the dialog design.

FIG. 12 is a flowchart (No. 1) showing the operation of the fourth mode according to the present invention, and FIG. 13 is the same flowchart (No. 2).

The feature of the fourth aspect is that, when transitioning to the second dialogue during the frame processing of the first dialogue, the dialogue data filled in the slot in the first dialogue is held as it is. Further, among the dialogue data to be filled in the slots in the frame processing in the above-mentioned second dialogue, those held data are diverted for those which overlap with the held dialogue data. Is.

Referring to FIG. 12, steps S31 and S32 are the same as steps S21 and S22 shown in FIG. 9, respectively.

Referring to FIG. 13, steps S33 and S
34, S35 and S36 are the same as steps S23, S24, S25 and S27 shown in FIG. 9, respectively.

Therefore, the feature of this fourth aspect is that FIG.
3 is represented by steps S37 and S38.

When transitioning to the second dialogue while frame processing is being performed for the first dialogue, and when the dialogue continues without ending, that is, when the result of determination in step S33 in FIG. 12 is No, The judgment result of step S35 in FIG. 13 is also No.

At this time, the dialogue data filled in the slot in the frame processing in the first dialogue is held as it is (step S37 in FIG. 13).

Further, among the dialogues to be filled in the slots in the frame processing in the above-mentioned second dialogue of the transition destination, for the dialogues that overlap the held dialogue data, the held dialogue data are diverted. That is, the frame data used at the transition destination is read from the frame data area 22 (step S38 in FIG. 13).

As described above, in the fourth mode, it is possible to reuse common dialogue data and thereby reduce the memory amount in the dialogue database 15.

FIG. 14 is a flow chart (No. 1) showing the operation of the special case of the fourth aspect of the present invention, FIG.
Is the same flowchart (No. 2).

The special case of the fourth mode defines the case where the "second dialog" in the above-described fourth mode is the assisted dialog described above (FIG. 11). Therefore, the "first dialogue" in the fourth mode is the main dialogue.

That is, in the special case of the fourth aspect, the voice recognition processing unit 11 (FIG. 1) determines that assistance to the user is required during frame processing for the main dialogue, and When transitioning to the assisted dialogue, the dialogue data filled in the slot in the main dialogue is held as it is, and among the dialogue data to be filled in the slot in the frame processing in the assisted dialogue, the held dialogue data described above is held. With respect to those overlapping with, the retained data is reused.

Referring to FIGS. 14 and 15, FIG.
Step S22 shown in FIG. 15 and Step S23 shown in FIG.
S27 is the same as steps S22 and S23 to S27 shown in FIG. 9, respectively.

Therefore, the notable steps in the special case of the fourth mode are steps S41 and S4 shown in FIG.
2, S43 and S44. However, the steps S43 and S44 are respectively performed by the fourth step shown in FIG.
This is the same as steps S37 and S38 in the above embodiment.

For this reason, steps S41 and S42 are particularly noteworthy steps.

When the user does not respond to the inquiry from the system side and a predetermined time (for example, 5 seconds) elapses (timeout), the determination result of the above step S41 becomes Yes, the transition to the assist dialogue starts, and The process proceeds to step S43, which is a feature of the fourth aspect, and further to step S44.

Even if the above-mentioned timeout does not occur, but the response from the user is for "help" and a transition to the assist dialogue is made, the step 43, which is the feature of the fourth aspect, is reached. , And then step S4
Up to 4.

As a result, the effect obtained in the fourth mode (by changing the use of the dialogue data, the dialogue database 1
The effect similar to (reducing the memory amount of 5) can be obtained during the transition to the assisted dialogue.

FIG. 16 is a transition diagram showing a dialogue model example I in the special case of the fourth mode, and FIG. 17 is a transition diagram showing a dialogue model example II in the special case of the fourth mode.
FIG. 13 is a transition diagram showing a dialogue model example III in the special case of the fourth mode.

In FIG. 16, in the above assist dialogue,
The case where the slot value attribute A described with reference to FIG. 3 and the slot value attribute B also described with reference to FIG.

The left side of FIG. 16 is the transition sequence of the main dialogue, and the right side thereof is the transition sequence of the assist dialogue.

Frame data acquired in the transition sequence of assist dialogue (slot value attribute A, slot value attribute B)
Is used as frame data in the frame processing of the main dialogue.

The view of FIGS. 17 and 18 is the same as the view of FIG. 6 described above.

FIG. 19 is a flowchart (No. 1) showing the operation of the fifth aspect of the present invention, and FIG. 20 is the same flowchart (No. 2).

The feature of this fifth mode is that it is shown in FIG.
18 is characterized in that, among a plurality of slots forming a frame, only the slot filled with the dialogue data retains the filled dialogue data.

As a result, the effect of reducing the memory amount of the dialogue database 15 by reusing the dialogue data, which is obtained in the modes of FIGS. 12 to 18, becomes even more remarkable.

Referring to FIGS. 19 and 20, first, in FIG. 19, steps S31, S32, S33 and S35 are similar to the steps of the same numbers shown in FIGS. 12 and 13 described above.

Next, in FIG. 20, steps S34 and S36 are the same as the steps of the same numbers shown in FIG. Therefore, steps S51 and S5
2 is a noteworthy step in the fifth aspect. However, regarding step S52, step S shown in FIG.
Similar to 27.

In step S51, among the plurality of slots forming the frame, only the slot filled with the dialogue data is held in the dialogue database 15 for the filled dialogue data.

Under each of the modes before the fifth mode, all the slots forming the frame of the frame data structure shown in FIG. 3 are held regardless of whether the slot is empty or not. I am trying.

However, in the fifth mode, the above step S5 is performed.
As shown in FIG. 1, only the result of slot value (value 1, value 2, etc.) and attribute are held. For example, only data indicating “Hyogo prefecture” and “destination setting” that fill the slot are held.

As a result, as described above, the memory amount can be further reduced.

FIG. 21 is a flowchart (No. 1) showing the operations of the sixth, seventh and eighth modes according to the present invention, and FIG. 22 is the same flowchart (No. 2).

First, the sixth mode will be described. In the sixth mode, the dialog data to be filled in the slot in the frame processing in the second dialog or the assist dialog at the transition destination and the dialog data held as described above. It is characterized in that it is determined whether or not and that the slot is filled by using the held dialogue data. The "held dialogue data" means step S37 in FIG. 13 or step S43 in FIG.
It is the data held by.

Referring to FIGS. 21 and 22, this sixth
This mode is represented by steps S61 → S62 → S63 → S64.

In FIGS. 13 and 14 described above, the processing ends in steps S38 and S44, respectively.
The sixth (and seventh and eighth) modes described here are modes after Steps S38 and S44. That is, the process starts from step S61 of FIG.

In step S62, the dialogue data to be filled in the slot in the frame processing of the transition destination, for example, the slot value attributes 1A, 1B, 2A, 2B ... (Refer to FIG. 3) and the dialogue data held above (this example). Then, it is determined whether or not the slot value attributes 1A, 1B, 2A, 2B ...) Match.

If they match, the corresponding slot in the frame at the transition destination is filled with the contents of the matched slot in step S63.

In this way, the frame reference table 2 has the frame pattern of the frame in which the slots are filled.
3 (FIG. 5), the conversation ID / conversation state ID of the next transition destination is searched, and the conversation is further advanced. This is step S64 in FIG.

As a result, also in the sixth aspect, the effect of reducing the amount of memory by reusing the interactive data can be obtained.

Subsequently, referring to FIGS. 21 and 22,
The seventh aspect will be described.

The seventh mode is whether or not the dialogue data to be filled in the slot in the frame processing in the second dialogue or the assist dialogue of the transition destination already described and the above-mentioned held dialogue data match. If it is determined that they do not match, and if it is determined that the current frame processing at the transition destination has already progressed from the start of frame processing, then return to that point and execute the reverse search of the conversation data. It is characterized by.

Referring to FIG. 21, the step S62 is executed.
If the result of the determination is No, that is, they do not match, it is first determined in step S63 'what the current conversation ID / conversation state ID is. Based on this determination, the process moves to step S65 in FIG.

In step S65, it is determined whether or not the current frame processing at the transition destination is in a state to start frame processing.

The seventh mode is characterized by this step S65 and the subsequent step S67. Preferably step S66 is incorporated. The determination result of step S65 is Y
If es, that is, if the current frame processing is the start of frame processing, the process proceeds to step S66 and the previously held conversation data is cleared. This is because the frame processing is started anew and the held conversation data becomes unnecessary.

On the other hand, the result of the determination in step S65 is N.
o, that is, when the current frame processing has already proceeded from the start of the frame processing, the necessary dialog data has already been filled in the slot group, and therefore the original processing is returned to the reverse search of the frame data. Up to this point is the seventh mode.

The eighth aspect then begins. In the eighth mode, the above reverse search is executed (step S67), and the dialogue data obtained there is used to fill the slot in the frame processing in the above-mentioned first dialogue or main dialogue (step S68). It is characterized by that.

Referring to FIG. 22, this eighth aspect is represented by steps S68 → S69 → S64 (previously described).

With respect to the frame pattern thus obtained, the conversation reference ID / conversation state ID of the transition destination is searched by the frame reference table 23, and the dialogue is further advanced in the same manner as in the first mode.

As a result, the seventh and eighth aspects also bring about the effect of reducing the memory amount.

FIG. 23 is a flow chart showing the operation of the ninth mode according to the present invention.

In the ninth aspect, when the type of dialogue at the transition destination is the same as the type of dialogue before the transition, all the dialogue data in the frame generated by the frame processing in the dialogue before the transition is cleared. Then, the frame processing in the dialogue before this transition is executed again.

In FIG. 23, in step S71, it is determined whether the dialogue at the transition destination is the same as the dialogue before the transition.

If they are the same (Yes), the process moves to step S72 to clear all the contents of the slot filled in by the frame processing before the transition.

After that, the frame processing before the transition is performed again from the beginning.

If the data of the slot of the frame before the transition is left as it is, the frame pattern (frame patterns 1, 2 ... In FIG. 5) specified by the frame still filled must be prepared separately. This is because it has to be eliminated.

Therefore, according to the ninth aspect, it is not necessary to repeatedly hold data of the same format, and the effect of reducing the memory amount is brought about.

FIG. 24 is a diagram showing an example of a dialogue model in which the ninth mode is used.

This is a dialogue S in navigation.
-5 is taken out.

Since the user's reply to S-5: 5 is "yes", the process returns to the same dialog S-5. When returning, S-5: 2 and S-5: 4 "No" and S
−5: 6 Clear “Yes” together.

FIG. 25 is a flow chart showing the operation of the tenth aspect of the present invention.

In the tenth aspect, the instruction content of the transition destination of the dialogue specified by the frame reference table 23 is
Application 13 with which the dialogue system 10 is associated
When it indicates that a branch should be made in accordance with the above, it is necessary to temporarily terminate the dialogue processing which has been proceeding until immediately before and continue the dialogue processing at the transition destination designated by the application 13 together with the dialogue start instruction. It is a feature.

Referring to FIG. 25, in step S81, the instruction content of the transition destination of the above dialogue is a specific control code, that is, a code indicating that the transition destination should be decided by an application (also referred to as an application) (Y
es), and proceeds to step S83. This control code can be written in the extension area 26 (FIG. 11).

If it is not a specific control code (N
o), the procedure moves to step S82, and the normal processing described above is performed.

If a specific control code is described, the process moves to step S83 to end the interactive process once. At this time, the API (Application) of "dialogue end"
Interface) is called.

Then, in step S84, the application waits for an instruction to start a dialogue.

When this dialog start command is issued, step S8
5, dialogue ID / state I sent with this command
The dialogue processing is started from the dialogue defined by D (transition destination designation data).

As described above, according to the tenth aspect, since the application for instructing various dialogue states issues the instruction by itself, the amount of data to be stored in the dialogue database 15 can be greatly reduced.

FIG. 26 is a diagram showing an example of a dialogue model to which the tenth aspect is applied.

This figure shows how the transition destination branches according to the instruction of the application.

FIG. 27 is a flow chart showing the operation of the eleventh aspect of the present invention.

In the eleventh aspect, a part of a series of slots (frame data area 22 in FIG. 3) is provided with dialogue data provided from the application 13 based on a request to the application 13 with which the dialogue system is associated. It is characterized by being filled by.

In the conventional navigation device, when the interactive data is input by the operation switch of the operation panel (see SW in the upper right of FIG. 1) instead of the voice, all the operation switches are operated until the series of operations is completed. It was not possible to have a dialogue without him.

However, according to the eleventh aspect, if the dialog data obtained by the switch operation according to the application, for example, U: “Hyogo prefecture” or U: “Kobe city” exists, this is used by the application 13 side. To be received from the application 13.

As a result, the dialogue data obtained by the switch operation can be managed by the dialogue transition control means 21, and the flexibility of dialogue design is enhanced.

Referring to FIG. 27, in step S91,
When the application 13 issues a "dialogue start" command, in step S92, the designated data of the transition destination output together with the command is analyzed to determine whether or not to start the dialog midway through the operation.

In step S93, if the decision result in step S92 is No, the application 13 is requested to provide a value necessary to fill a slot that has not been filled by the time the current interactive state is reached.
Then, the required value is filled in the corresponding slot.

In step S95, it is determined whether or not the frame pattern of the frame data filled with the acquired value exists in the frame reference table 23, and if there is, the dialog is continued in step S96.

FIG. 28 is a flow chart showing the operation of the twelfth aspect of the present invention.

In the twelfth aspect, when an interruption occurs in the application 13 with which the dialogue system 10 is linked while the voice synthesizing unit 16 is activated after a certain dialogue state S to guide the user, The dialogue transition control means 21 controls the dialogue state S by the application 13.
It is characterized in that the destination of transition is specified to the dialogue state S in response to the instruction to return to.

In this case, it is preferable to clear the frame data accumulated in the dialogue state S.

By transferring a part of the control of the state transition to the application 13 side as described above, it is possible to reduce the load on the dialogue engine (dialog transition control means 21). That is, it is not necessary to prepare frame data (slot) for returning.

Referring to FIG. 28, step S101
Then, the dialogue engine 12 determines whether or not an interrupt has occurred in the application 13. As an example, when it is determined that the voice recognition processing unit 11 has performed erroneous recognition, the interrupt is generated. Then, in step S102, it is determined whether a command to return to the previous dialogue has been issued. If yes, step S
At 103, the dialogue transition control means 21 returns to the frame that forms the dialogue that led to the above-mentioned erroneous recognition, and reconstructs the frame.

FIG. 29 is a diagram showing a thirteenth aspect of the present invention.

The thirteenth mode has a configuration in which a global speech dedicated frame memory 31 is added to the configuration of FIG.

Among the various dialogues described above, there is always an utterance common to all dialogues. This is global utterance. Some examples are utterances input by the user due to “return”, “recognition error”, “stop”, “temporary stop”, “misrecognition”, “help”, and the like.

If the frame data area 22 and the frame reference table 23 are formed including such an utterance, the data size becomes large. Therefore, the amount of memory increases. Also, the maintenance described above becomes difficult.

Therefore, the thirteenth aspect is characterized in that a global utterance memory 31 for exclusively storing only the above-mentioned global utterance is provided.

Thus, it is possible to obtain the effect that the memory amount can be reduced and the maintenance can be improved.

[0169]

As described above, according to the present invention, all dialogues can be realized by frame processing, thereby realizing a dialogue system in which the degree of freedom, flexibility and simplicity of dialogue design are improved. You can

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a dialogue system to which the present invention is applied.

FIG. 2 is a diagram showing a main part of the present invention.

3 is a diagram showing a detailed example of a frame data area 22 of FIG.

FIG. 4 is a diagram showing an example of a dialogue ID and a dialogue state ID.

5 is a diagram showing a detailed example of a frame reference table 23 of FIG.

6 is a diagram schematically showing how to search for the frame pattern of FIG.

FIG. 7 is a flowchart showing the operation of the first aspect according to the present invention.

8A and 8B are transition diagrams showing two simple examples of a dialogue model according to the first aspect.

FIG. 9 is a flowchart showing the operation of the second aspect according to the present invention.

FIG. 10 is a transition diagram showing an example of a dialogue model according to a second aspect.

FIG. 11 is a diagram showing an operation according to the third aspect of the present invention.

FIG. 12 is a flowchart (part 1) showing the operation of the fourth aspect according to the present invention.

FIG. 13 is a flowchart (part 2) showing the operation of the fourth aspect according to the present invention.

FIG. 14 is a flowchart (No. 1) showing the operation of the special case of the fourth aspect according to the present invention.

FIG. 15 is a flowchart (part 2) showing the operation of the special case of the fourth aspect according to the present invention.

FIG. 16 is a transition diagram showing a dialogue model example I in the special case of the fourth mode.

FIG. 17 is a transition diagram showing a dialogue model example II in the special case of the fourth mode.

FIG. 18 is a dialogue model example III in the special case of the fourth mode.
It is a transition diagram showing.

FIG. 19 is a flowchart (No. 1) showing the operation of the fifth aspect according to the present invention.

FIG. 20 is a flowchart (No. 2) showing the operation of the fifth aspect according to the present invention.

FIG. 21 is a flowchart (No. 1) showing the operations of the sixth, seventh, and eighth aspects of the present invention.

FIG. 22 is a flowchart (No. 1) showing the operations of the sixth, seventh and eighth aspects of the present invention.

FIG. 23 is a flowchart showing the operation of the ninth aspect according to the present invention.

FIG. 24 is a diagram showing an example of a dialogue model in which the ninth mode is used.

FIG. 25 is a flowchart showing an operation according to the tenth aspect of the present invention.

FIG. 26 is a diagram showing an example of a dialogue model in which the tenth aspect is used.

FIG. 27 is a flowchart showing the operation of the eleventh aspect according to the present invention.

FIG. 28 is a flowchart showing the operation of the twelfth aspect of the present invention.

FIG. 29 is a diagram showing an operation according to the thirteenth aspect of the present invention.

[Explanation of symbols]

10 ... Dialog system 11 ... Voice recognition processing unit 12 ... Dialogue engine 13 ... Application 14 ... Recognition dictionary 15 ... Dialogue database 16 ... Voice synthesizer 17 ... GUI control unit 18 ... Display 21 ... Dialogue transition control means 22 ... Frame data area 23 ... Frame reference table 25 ... Transition destination ID area 26 ... Extended area 31 ... Global speech memory

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Hideki Kitao             1-228 Goshodori, Hyogo-ku, Kobe-shi, Hyogo               Within Fujitsu Ten Limited (72) Inventor Hideki Nakamura             1-228 Goshodori, Hyogo-ku, Kobe-shi, Hyogo               Within Fujitsu Ten Limited F-term (reference) 5D015 HH00 KK02 KK04 LL10 LL12                 5D045 AB17

Claims (15)

[Claims] 1. A voice recognition processing unit for recognizing the content of a voice uttered by a user, a dialogue engine for generating a guidance to a user according to an application according to the recognition result, and a synthesized content of the generated guidance as a voice. In a dialogue system comprising at least a voice synthesis unit provided to a user and a dialogue database which cooperates with the dialogue engine to generate the guidance, a series of slots forming frames constituting a dialogue are sequentially arranged. A dialogue system characterized by comprising dialogue transition control means for specifying a transition destination of a dialogue according to a filling degree of the dialogue data to be filled. 2. A voice recognition processing unit for recognizing the content of a voice uttered by a user, a dialogue engine for generating a guide to the user according to an application according to the recognition result, and a synthesized content of the generated guide as a voice. In a dialogue system comprising at least a voice synthesis unit provided to a user and a dialogue database that cooperates with the dialogue engine to generate the guidance, each of a series of slots forming a frame that constitutes a dialogue. Frame data area for filling the corresponding dialogue data in the corresponding slot, and a plurality of frame patterns corresponding to each of the plurality of frame patterns determined by the filling degree of the dialogue data filled in the frame data area. A frame reference table that defines the transition destination for each frame pattern, and Forming a dialogue transition control means in the dialogue engine, which is constructed in the base, and uses the frame data area and the frame reference table to specify the transition destination of the dialogue according to the filling degree of the dialogue data. A dialogue system characterized by. 3. When transitioning to a dialogue ID different from the dialogue ID of a dialogue in progress before each dialogue data is completely filled in all of the series of slots forming the frame, the frame processing that has been progressing so far is terminated. The dialogue system according to claim 2, wherein 4. Each of the frame patterns in the frame reference table has a transition destination ID area consisting of a transition destination conversation ID and a transition destination conversation state ID defined for each frame pattern, and also for each frame pattern. 3. The interactive system according to claim 2, further comprising an extension area that describes a transition destination to a specific event defined in 1. 5. When transitioning to a second dialogue during frame processing for the first dialogue, the dialogue data filled in the slot in the first dialogue is held as it is, and The dialogue data to be filled in the slot in the frame processing in the dialogue 2 is duplicated with the held dialogue data, and the held data is used. Dialog system. 6. The voice recognition processing unit determines that assistance to the user is required during frame processing for the main dialogue, and when transitioning to the assistance dialogue due to the determination, the voice recognition processing unit performs the above-mentioned operation in the main dialogue. Of the dialogue data to be held in the slot as it is and to be filled in the slot in the frame processing in the assist dialogue, the held data is the same as the held dialogue data. 3. The dialogue system according to claim 2, wherein the dialogue system is diverted. 7. The dialogue according to claim 5, wherein only the slot filled with the dialogue data among the plurality of slots forming the frame holds the filled dialogue data. system. 8. It is determined whether or not the dialogue data to be filled in the slot and the held dialogue data in the frame processing in the transition destination second dialogue or assist dialogue are coincident with each other, and if they are coincident with each other. The dialogue system according to claim 5 or 6, wherein when the determination is made, the held dialogue data is used to fill the slot. 9. It is determined whether or not the dialogue data to be filled in the slot and the held dialogue data in the frame processing in the transition destination second dialogue or assist dialogue does not match. When it is determined that the current frame processing of the transition destination has already progressed from the start of the frame processing, the process returns to the original state and the reverse search of the conversation data is executed. Item 7. A dialogue system according to item 5 or 6. 10. The dialogue according to claim 5, wherein the reverse search is executed and the dialogue data obtained there is used to fill a slot in frame processing in the first dialogue or the main dialogue. system. 11. When the type of dialogue at the transition destination is the same as the type of dialogue before the transition, all the dialogue data in the frame generated by the frame processing in the dialogue before the transition are cleared and then the transition is performed again. The dialogue system according to claim 2, wherein the frame processing in the previous dialogue is executed. 12. When the specified instruction destination of the transition destination of the dialogue indicates that the dialogue system should branch in accordance with the application with which the dialogue system is linked, the dialogue processing that has been proceeding to the immediately preceding time is terminated. 2. The dialogue system according to claim 1, wherein the dialogue processing is continued at a transition destination designated by the application together with a dialogue start instruction. 13. The dialogue system according to claim 1, wherein a part of the series of slots is filled with dialogue data provided by the dialogue system based on a request to the application associated with the dialogue system. . 14. If an application associated with the dialogue system is interrupted while the speech synthesizer is being activated after a certain dialogue state, the dialogue transition control means changes from the application to the dialogue state. The dialogue system according to claim 1, wherein a transition destination is specified to the dialogue state in response to an instruction to return. 15. The dialogue system according to claim 1, further comprising a global utterance memory that exclusively stores global utterances that are commonly used in various kinds of dialogues.