JPH03233677A - Conversation control system - Google Patents

Abstract

PURPOSE:To find the flow error of conversations by providing a scene with an antecedent by which the scene is true and a consequent by which the scene is terminated and checking the antecedent and the consequent of the scene as change elements with respect to the change of a stack structure. CONSTITUTION:A flow error finding part 12 compares a present stack 17, which is used to trace the flow of conversations by a conversation control part 11 and indicates the flow of conversations corresponding to the current input, and a preceding stack 18 indicating the flow of conversations corresponding to the preceding input with each other and discriminates whether the antecedent and the consequent of the scene are satisfied or not to find the flow error of conversations. When finding the flow error of conversations, the part 12 transfers speaking contents indicating the flow error of conversations to an output sentence generating part 13 and returns the state of the present stack 17 to that of the preceding stack 18. But otherwise, conversation contents of a system determined in the conversation control part 11 are transferred to the output sentence generating part 13. Thus, the flow error of conversations is found.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a conversation control method for a computer system having a human-machine interface using natural language. The aim is to discover things that do not follow the flow of the conversation that took place.

[Prior Art] In a computer system having a human-machine interface using natural language, in order to determine the computer's utterances, it is necessary not only to understand the user's utterances but also to understand the flow of the conversation. Conventionally, there have been the following methods for computers to understand the flow of conversation during a conversation.

As a first method, a method has been proposed in which the semantic relationship between successive utterances between the computer and the user, or between consecutive utterances by the user, is extracted from grammatical information such as conjunctions and adverbs. (Reference 1 "Introduction to Cognitive Science", by Masashi Toda et al., Science Publishing, pp. 20B-216,
(1986). However, with this method, although it is possible to understand the semantic relationship between two sentences, it is not possible to trace the flow of the entire conversation. In other words, with this method, only a very general flow can be understood.

Second, using a script consisting of a sequence of events (scenes) that would typically occur in a certain place or event,
A method of understanding the flow of a conversation by storing a history of which events on the script have been talked about in the conversation up to now, and by associating the current user's utterance with a certain event on the script 1. has been proposed (Reference 2, “Artificial Intelligence Handbook Volume 1”, Kyoritsu Shuppan, pp. 389-393. 1983)
. However, in this method, the flow of the conversation is statically described in a script, and it is not possible to understand unusual jumps in the flow of the conversation.

As a method to overcome the drawbacks of the first and second methods,
The third step is to express (understand) the flow of the conversation up to the present in a stack or tree structure that is a structured structure of multiple scenes, using multiple scenes that can be the focus of a conversation and the knowledge to dynamically assemble the scenes. The following method has been proposed (Reference 3 "Dialogue Management Function of Intelligent Dialogue System", Hitomi Shimada et al., ICOT Research Bulletin TH-0162, 1986).

In this third method, the focus of the current conversation is the top of the stack or leaf of the tree pointed to by the focus pointer, and the stack or tree represents how the current focus has emerged from the flow of the conversation so far. There is. When a user or computer utters an utterance, the utterance is associated with a specific scene, and using the knowledge to dynamically assemble the scene, the scene associated with the utterance and the scene in the stack or tree are associated with the scene. For example, if a new scene is generated by a user's utterance, that scene is placed at the top of the stack, or the new scene is combined below the leaf pointed to by the focal point or ink. Thus, the third method manipulates the stack or tree to control the focus of the current conversation.

[Problem to be Solved by the Invention] In all of the three methods described above, including the third method that can solve the disadvantages of the first and second methods, the computer tracks the flow of the conversation created by the user. We understand the flow of conversations in order to conduct conversations that reflect the user's intentions.
It is assumed that there are no errors in the flow of conversation by the user.

However, considering the flow of the conversation, it is inevitable that the wrong user will engage in a conversation. The three methods described above do not detect conversational flow errors.

An example of a computer system having a human-machine interface using natural language is a language training system. In such a language training system,
The learner's utterances unduly disrupt the flow of the conversation at that point because the learner misunderstands the system's utterances or the flow of the conversation, or does not accurately understand the system's utterances or the flow of the conversation. (Reference 4: “Configuration of intelligent CAI for language training based on conversation simulation J”
, by Hideki Yamamoto et al., Transactions of the Information Processing Society of Japan Vol. 30, No. 7,
pp. 908-917.1989). In this case,
The system understands when the learner disrupts the flow of the conversation and
When the learner makes a mistake, the student must point out the error and take steps to restore the correct flow of the conversation.

It is not appropriate to independently define a method for understanding and controlling the flow of a conversation and a method for discovering errors in the flow of the conversation from the standpoint of improving processing efficiency. Therefore, it is conceivable to use the above-described method of understanding and controlling the flow of conversation to detect errors in the flow of conversation.

In a second method using scripts as described above, the user disrupts the flow of the conversation when the computer system is no longer able to associate the user's input with events on the script before detecting the end of a certain script. It can be considered as However, a third method that dynamically generates and manages conversation flow using a stack structure or tree structure
In this method, it is not possible to directly discover what kind of state is an error in the flow of a user's conversation.

Note that a method for detecting errors in the flow of conversation using the second method has been separately proposed by the same applicant.

The present invention has been made in consideration of the above points, and uses a method that dynamically generates and manages the flow of a conversation using a stack structure or a tree structure. The purpose of the present invention is to provide a conversation control method that can detect errors in the flow of conversation caused by unduly disrupting the flow of conversation.

[Means for Solving the Problems] In order to solve the problems, the present invention provides a conversation control method for a computer system having a human-machine interface using natural language, in which the flow of conversation is adjusted for each scene. We developed a conversation control method that generates and manages the conversation flow by dividing the scene and dynamically constructing the scene as a stack structure or a tree structure, as follows.

In other words, a scene has a precondition for the scene to hold and a postcondition for the scene to end, and the preconditions and postconditions for the scene corresponding to changes in the stack structure or tree structure are examined. I decided to discover errors in the flow of the conversation.

[Operation] The present invention divides the flow of conversation into scenes for each scene,
It is assumed that a method is used to generate and manage the flow of conversation by dynamically constructing scenes as a stack structure or tree structure.

In order to transition to a scene with a conversation, the conversation that actually took place before that scene must be specified to some extent, and a predetermined scene or conversation must exist before (not necessarily immediately before) that scene. . Therefore, it can be said that there are specific conditions for entering that scene. Therefore, in the present invention, scene information for constructing a stack structure or a tree structure is provided with preconditions.

Also, in order to be able to remove a scene from a certain conversation, there are actually some conversations and scenes that must be carried out at least following that scene. That is, there are conditions for ending the scene.

Therefore, in the present invention, scene information for constructing a stack structure or a tree structure is provided with a post-condition.

Then, when the stack structure or tree structure changed, we decided to discover errors in the flow of the conversation by examining the preconditions and postconditions of the changing elements of the scene.

“Example” An example of the present invention will be described in detail below with reference to the drawings.

Here, FIG. 1 is a functional block diagram of a computer system to which this embodiment is applied, FIG. 2 is an explanatory diagram showing an example of a meaning expression that is the output of the input sentence understanding unit 10, and FIG. FIG. 4 is an explanatory diagram showing an example of a scene used for conversation control. FIG. 5 is an explanatory diagram showing an example of knowledge for inferring a specific scene from a user's utterance. is a processing flowchart of the conversation control section 11, and FIG. 7 is a processing flowchart of the conversation flow error detection section 12.

Although not shown, like other computer systems, this system also includes a central processing unit (CPU), a memory, a display device, a keyboard device, and a hard disk device.

The hard disk drive contains natural language processing programs for understanding input sentences and generating output sentences, conversation control programs for determining output sentences according to input sentences, processing errors in conversation flow, etc., as well as programs for processing errors at each stage. Various rules and knowledge used for processing are stored. For example, a program for detecting errors in the scene shown in FIG. 4 and the flow of conversation shown in FIG. 7, which will be described later, is stored in the hard disk device. Natural language processing programs and conversation control programs are loaded into memory and executed by the CPU.

Note that, as is well known, the display device and the keyboard device are for a man-machine interface between the user and the system, and dialogue is performed through these devices.

First, the functional configuration of this computer system will be explained using FIG.

In FIG. 1, from the input of the user's utterance in natural language until the utterance of the system in natural language, there are mainly processing by the input sentence understanding unit 10, processing by the conversation control unit 11, and flow errors. Processing is executed in the order of processing by the discovery unit 12 and processing by the output sentence generation unit 13.

After the user speaks, the input sentence understanding unit 10 converts the natural language input into a format called a semantic expression that can be processed (inferred) by the system. It is set in the memory 14.

FIG. 2 shows an example of a semantic expression. This semantic expression has a pair of slot name 20 and slot value 21, and the meaning 22 of the value is also shown in FIG. 2 for reference. Figure 2 shows "I want t.

This is an example of the meaning of the sentence "payl) y card."

In the case of this sentence, for example, Slowa (−
The slot with the name “prediCate” has the value “pay”.
” is inserted, and the slot name ゛ag which means “actor” is inserted.
The value "'user" is inserted into the slot of "ent".

After the semantic expression corresponding to the user's input sentence is set in the situation memory 14, the conversation control unit 11 tracks the flow of the conversation and performs processing to determine the content of the system's utterance. The conversation control unit 11 includes a scene set 15 that is knowledge representing the flow of the conversation, scene inference knowledge 16 that infers a specific scene from the user's utterances, a situation memory 14, and a current memory that controls the focus of the conversation. Four pieces of information in the stack 17 are used.

The flow error detection unit 12 uses a current stack 17 representing the conversation flow corresponding to the current input, which is used by the conversation control unit 11 to track the conversation flow, and a previous stack 17 representing the conversation flow corresponding to the previous input. By comparing the stack 18 of the scene and determining the sufficiency of the pre-condition and post-condition of the scene, a conversation flow error is discovered, and when a conversation flow error is discovered, an utterance indicating a conversation flow error is detected. The contents are passed to the output sentence generation unit 13, and the current state of the stack 17 is returned to the previous state of the stack 18. In other cases, the system utterance content determined by the conversation control unit 11 is passed to the output sentence generation unit 13.

The output sentence generation unit 13 generates a natural language sentence (output sentence) to be actually output according to the utterance content determined by the conversation control unit 11 or the flow error detection unit 12.

Next, the scene set 15, scene inference knowledge 16, situation memory 14, and stacks 17 and 18 used by the conversation control unit 11 will be sequentially explained using the drawings.

Each of the scenes 15a to 15d included in the scene set 15 has a scene name 30 for identifying the scene and a name for the scene to be established, as shown in FIGS. 4(A) to 4(D). It has a necessary precondition 31, a rule group 33 that is a set of rules for obtaining certain information from a conversation in that scene, and a postcondition 32 that is satisfied as a result of applying the rules in that scene.

In order to move to a certain scene of conversation, the conversation that actually took place before that scene must be specified to some extent, and a predetermined scene (conversation) must exist before (not necessarily immediately before) that scene. There are some scenes in which this happens, and the precondition 31 is provided in consideration of this point.

Furthermore, in order for a conversation to be extracted from a certain scene, there are actually conversations and scenes that must be completed at least in that scene, and the post-condition 32 is provided in consideration of this point.

In the pre-condition 31 and the post-condition 32, a semantic expression in the same form as the semantic expression in the situation memory 14 is described. Each rule in the rule group 33 has a rule number 34 and a condition part 35.
and an execution section 36, and the condition section 35 includes a situation memory 1.
A semantic expression expressed in the same form as the semantic expression in 4 is written, and the execution unit 36 sets the semantic expression representing the native language into the situation memory 14, which is obtained from the system's utterance, calling a new scene, or having a conversation. is described.

The scene inference knowledge 16 used by the conversation control unit 11 will be explained using FIG. 5. The scene inference knowledge 16 consists of a knowledge number 40, a condition part 41, and a scene part 42, as shown in FIG. The condition part 41 contains the state of the semantic expression obtained from the user's input sentence for firing, or
The state of the semantic expression in the situation memory 14 at that point in time is described, and the scene section 42 describes the scene name of the scene that will be recalled when the condition is met.

The situation memory 14 is a memory that holds conversation history, and holds semantic expressions obtained from conversations between the user and the system that have taken place up to now. It also holds a semantic expression indicating the outcome obtained as a result of the conversation, which is set in the situation memory 14 by the execution of the rule.

The current stack 17 is used to control the focus of the conversation. As shown in FIG. 3, scenes activated during the conversation are placed on a stack 17. The most recently activated scene is at the top of the current stack 17. Only for the scene at the head of the current stack 17, it is determined whether or not to fire a rule related to that scene. The previous stack 18 contains the scene contents of the stack 17 at the time of system utterance.

Next, the processing of the conversation control section 11 will be explained using the flowchart shown in FIG. Note that FIG. 6 shows how the stack 17 is operated, and such processing will be explained.

First, it is determined whether the internal meaning expression currently set in the situation memory 14 satisfies the postcondition of the scene at the top of the stack 17 (step 50).

If the condition is satisfied, it is OK to extract from this scene, so the first scene is removed from the current stack 17 (step 51).

When such a scene removal process is executed or when the post-conditions are not satisfied, a new scene is created using the internal semantic representation set this time in the situation memory 14 and the knowledge for inferring a specific scene from the user's utterances. It is determined whether or not inference can be made (step 52). If you can't infer,
The process immediately proceeds to step 56 and subsequent steps. On the other hand, if it can be inferred, it is further determined whether or not the new scene is on the stack 17, and if not, the new scene is placed on the stack 17, and if there is, the new scene is placed at the top of the stack 17, and the new scene is placed on the stack 17. After eliminating the scene, the focus of the conversation established by the user is set, and the process proceeds to step 56 and subsequent steps (steps 53 to 55).

In step 56, it is determined whether or not among the rules of the scene at the head of the current stack 17, there is a rule whose internal representation set in the situation memory 14 satisfies the firing condition.

If there is no such rule, after removing the first scene from the current stack 17,
(Step 58)
．． 59). If it exists, the focus of the conversation has changed, so the above-described determination process for the steering knob 56 is redone. If it does not exist, the operation of the stack 17 for the current utterance by the user is terminated.

Among the rules of the scene at the head of the current stack 17, if there is even one rule whose internal expression that has been sewn into the situation memory 14 satisfies its firing condition, the number of rules that can be fired is set to the number of rules parameter i. is set, and the initial value 0 is set to the rule parameter j (step 57). In addition, at this time, the number 0.1 is assigned to one or more rules that can be fired.
, . . . are associated with each other, and the parameter j specifies the rule by specifying one of the associated numbers.

Thereafter, it is confirmed by comparing the magnitudes of parameters i and j that processing for all rules that can be fired has not been completed (step 60).

When the processing for all the rules that can be fired is completed, the process proceeds to a process (step 56) of determining whether there is any rule that can be fired among the rules of the scene at the head of the current stack 17 mentioned above.

If processing for all rules has not been completed, the system utterance may be determined by firing the rule specified by parameter j, or the utterance of a new scene may be determined), then the parameter j is incremented and the next firing is performed. The possible rules are changed to those that indicate (step 61).

And, due to this ignition, a new scene is stacked 17
If the scene is stacked, the process returns to step 54 and it is determined whether or not there is a rule that can be fired for the stacked scene. Determine whether the length of the system utterance including the utterance determined by this firing is sufficient (
Step 62.63). If the length of the system utterance is not sufficient, the process returns to step 60 to determine whether processing of all rules that can be fired has been completed, and if it is sufficient, the stack operation processing for the current user's utterance is terminated.

The processing shown in FIG. 6 can be summarized and described as follows.

If the post-conditions for a scene on the stack 17 are satisfied by the processing in steps 50 and 51, that scene and the scenes stacked above that scene are removed.

If a certain scene can be inferred from the user's utterance through the processing of steps 52 to 55, that scene is placed on the stack 17.

Through the processing of steps 56 to 63, a rule is fired when the semantic expression corresponding to the user's input sentence satisfies the condition of the scene at the top of the stack 17, and the execution part of the rule is executed.

When another scene is activated as a result of the execution of the execution part, the focus of the conversation changes and another scene is placed on the stack 17. Also, during such processing, if there is no rule that can be fired in the active scene,
The first scene is removed from the stack 17, and the focus of the conversation returns to the scene when the removed scene was activated.

The processing by the conversation control unit 11 ends when the length of the system utterance determined here is sufficient or when there is no scene in the stack 17, and after the processing is finished, the flow error detection unit 12 will move.

Therefore, there are two cases (1) and (2) below when scenes are stacked on the stack through processing by the conversation control unit 11.

(1) When another scene is called from the current scene by applying the rule (step 6l-62-56) (2) When the scene inferred from the user's utterance is stack 1
If it is a new scene that is not on 7 (step 52-5
3-54> Furthermore, the scene is removed from the stack 17 in the following cases (3) to (5).

(3) If the post-condition of the scene holds (step 5
0-51> (4) All rules of the scene cannot be fired,
(5) If the scene inferred by the user's utterance is a previous scene on the stack 17 (Step 52-5)
3-55> Among cases involving such operations on the stack 17, cases (1) and (3) are definitely appropriate as changes in the flow of the conversation, but cases (2), (4) and ( 5) is sometimes not appropriate.

In case (2), if the natural flow of conversation is not satisfied between the new scene and the immediately preceding scene, it means that the user does not understand the flow of conversation in the immediately preceding scene. In other words, changing a scene to a new scene may be a good thing in itself, but there is a problem if the new scene is not appropriate in terms of the flow of the conversation.

In this case, if the preconditions for entering a newly accumulated scene are met, it can be said that the flow of the conversation is natural.

In case (4), if a scene with a matching rule exists in the stack 17, the focus of the conversation moves to that scene. It is also possible to return to a scene that was previously discussed. Therefore, stack 1 will be added immediately after the previous scene that was moved.
If the condition after the scene stacked in stack 17 is met, then the change in the flow of the conversation is appropriate, but if it is not the case, the user understands the scene stacked in stack 17 after the moved scene. It means you haven't done it.

In case (5), if the condition after the scene placed on the stack 17 immediately after the inferred scene (which in this case is equivalent to a scene in which a conversation has already taken place previously) is met, then the flow of the conversation is appropriate, but otherwise the user does not understand the scene placed on the stack 17 after the inferred scene.

Therefore, there are two possibilities that the flow of the conversation is incorrect. First, the condition after the scene is not met and the stack 17
and secondly, when the scene is removed from stack 1
This is the case when the precondition for the new scene does not hold even though a new scene has been stacked on 7.

The process executed by the flow error detection unit 12, the procedure of which has been determined based on the above consideration, will be explained using FIG. 7.

First, it is determined whether the stack 17 has changed from the previous stack 18 (step 70). If there is no change, it is determined that there is no error in the flow of the conversation since the conversations are in the same scene, and the error detection process is terminated.

If so, it is further determined whether the current stack 17 is the previous stack 18 with the scene removed (step 71). If so, it is determined whether there is any internal representation in the situation memory 14 that satisfies the post-condition of the removed scene (step 72).

If this determination is not made, the scene is ended and the scene is moved to another scene even though it is not possible, so it is determined that the flow of the conversation is incorrect and the discovery process is ended.

If the current stack 17 is not the one obtained by removing the scene from the previous stack 18, and if there is an internal representation in the situation memory 14 that satisfies the condition after the removed scene, then , it is determined whether the current stack 17 has scenes stacked on the previous stack 18 (step 73). If a negative result is obtained, it is determined that there is no error in the flow of the conversation, and the discovery process is terminated.

If scenes are accumulated, situation memory 1
It is determined whether any of the internal representations in 4 satisfy the precondition of the removed scene (step 74). If the result is that there is no error in the flow of the conversation, it is determined that there is an error in the flow of the conversation and the discovery process is terminated, and if the result is that there is an error in the flow of the conversation, it is determined that there is no error in the flow of the conversation and the discovery process is terminated.

The specific process flow will be explained below using FIGS. 3 to 7. Note that the conversation is between a hotel customer (user) and a clerk (system).

The previous stack 18 shown in FIG.
- This shows that the room scene has been called from the in scene. The current stack 17 also has the same contents before conversation control for the user's utterances is started.

In this state, the user says “That’s good, I
think I'If take the root"
(utterance a) and “I wander to keep
Assume that the following utterance is made: a+y baooaaes, (utterance b).

In this case, in response to utterance a, roo as shown in FIG.
In order to set a semantic expression equal to the post-condition of the m-scene in the situation memory 14, the post-condition of the room scene is satisfied and the room-t scene is removed from the stack 17;
The top of the stack 17 is check- shown in FIG. 4(A).
It becomes an in scene (step 5O-51).

Next, in response to the utterance, scene inference knowledge 1 shown in Figure 5
By using the knowledge of 6 whose knowledge number 40 is "1",
The keep-baggage scene shown in FIG. 4 (D> is loaded onto the stack 17 (steps 52-53-54>
.

The current stack in FIG. 3(B) shows the state of the stack 17 in that case. In the keep-baggage scene at the top of the stack 17, the rule whose rule number 34 is "0" can fire, so this rule fires and the 5peak brino-baagag
Execute e (steps 56-57-6O-61). As a result, if the length of the system's utterance is sufficient, the processing of the conversation control unit 11 ends (steps 62-63>).

In the flow error detection unit 12, as shown in FIG.
In this example, the previous stack 71 of FIG.
8 and the current stack 17 of FIG. 3(B).

Here, stack 17 has changed compared to previous stack 18 (affirmative result at step 70), and stack 17
Since the room scene has been removed from the stack 17 (affirmative result in step 71), the ro scene has been removed from the stack 17.
It is determined whether the post-om scene condition is satisfied by checking the semantic expression in the situation memory 14 (step 7).
2). At this time, since the semantic expression set in the situation memory 14 by the utterance a satisfies the postcondition of the room scene, an affirmative result is obtained.

Next, since a new scene has been placed on the stack 17 (affirmative result in step 73), the scene is placed on the stack 17 and it is determined whether the precondition for the scene is satisfied or not. It is determined by examining the semantic representation in the memory 14 (step 74).keep-bag
The preconditions of the cage scene correspond to the preconditions of the room scene and the postconditions that are known to be satisfied.
Since the 00m scene has been loaded onto the stack 17, it is clear that the precondition for the room scene holds true. Therefore, it is clear that the precondition for the keel)-baggage scene holds true, and a positive result is obtained in step 74. ), as a result, it turns out that the user did not make any conversation flow errors.

Next, assume that the previous stack 18 is in the state shown in FIG. 3 (A>) and the user only speaks.

In that case, the condition after the first roo+n scene is not satisfied and the room scene is not removed from the stack 17, but in response to the utterance, the scene inference knowledge 16 shown in FIG.
By using the knowledge whose knowledge number 40 is "1", k
The eep-bag(Jage scene is stacked at the top of the stack 17 (step 5O-52-53-54).

The current stack in FIG. 3(C) shows the state of the stack 17 in that case. keeD-bagCJa at the top of stack 17 (because rule number 0 can fire in the Je scene, rule number 0 fires,
5Execute peak bring-baggage (
Steps 56-57-60-61>. As a result, if the length of the system utterance is sufficient, the processing of the conversation control unit 11 ends (steps 62-63>).

In the flow error detection unit 12, as shown in FIG.
In this example, the previous stack 18 of FIG. 3(A) and the current stack 17 of FIG. 3(C) are compared (step 70).

Here, the stack 17 has changed from the previous stack 18, there is no scene removed from the stack 17, and a new keep-baggage scene is placed on the stack 17, so the keep-baggage scene placed on the stack 17
It is determined whether the preconditions of the p-baggage scene are satisfied by checking the semantic expression in the situation memory 14 (steps 71-73-74>).

Among the preconditions of the keep-baggage scene, ro
The condition corresponding to the post-condition of the om scene does not hold because utterance a is not made (negative result in step 74).

Therefore, it turns out that the user is making a mistake in the flow of the conversation.

Therefore, according to the above embodiment, a scene has a precondition for the scene to be established and a postcondition for the scene to end, and when the stack structure changes, the By examining the conditions and post-conditions, errors in the flow of the conversation can be discovered.

Therefore, it is possible to respond flexibly to errors in the user's conversation flow, such as pointing out the error in the conversation flow and maintaining the correct conversation flow.

Additionally, since flow errors are detected using a method for understanding the flow of conversation, processing is easier than when the understanding method and flow error detection method are provided separately. .

In addition, in the above-mentioned embodiment, the flow state of a conversation scene is dynamically expressed using a stack structure, but the present invention can also be applied to something that is dynamically expressed using a tree structure. be able to.

Further, the present invention can be applied not only to dialogue in English, but also to natural language interfaces in other languages, and the system is not limited to educational systems.

[Effects of the Invention] As described above, according to the present invention, a method of dynamically generating and managing the conversation flow using stacks and trees is used to prevent the user's utterances from inappropriately disrupting the flow of the conversation at that point. You can discover errors in the flow of conversation caused by disrupting the conversation.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram of a computer system to which an embodiment of the conversation control method according to the present invention is applied, FIG. 2 is an explanatory diagram showing an example of a semantic expression output from the input sentence understanding unit 10, and FIG. is an explanatory diagram showing an example of a stack representing the focus of a conversation, FIG. 4 is an explanatory diagram showing an example of a scene used for conversation control, and FIG. 5 is an explanatory diagram showing an example of knowledge for inferring a specific scene from a user's utterance. 6 is a processing flowchart of the conversation control unit 11, and FIG. 7 is a conversation flow error detection unit 12.
It is a processing flowchart. 10... Input sentence understanding unit, 11... Conversation control unit, 12
...Flow error detection unit, 13...Output sentence generation unit, 14
...Situation memory, 15...Scene collection, 16...
・Scene inference knowledge, 17...Current stack, 18・
...Previous stack, 31...Scene precondition, 32
...The post-scene condition.

Claims (1)

[Claims] A conversation control method for a computer system having a human-machine interface using natural language, which divides the conversation flow into scenes for each scene and dynamically constructs the scenes as a stack structure or a tree structure. In a conversation control method that uses a method to generate and manage the flow of conversation, a scene has a precondition for the scene to be established and a postcondition for the scene to end, and a stack structure or A conversation control method characterized by detecting errors in the flow of conversation by examining the preconditions and postconditions of scenes corresponding to changes in the tree structure.