JPH0612445A - Interactive processing controller - Google Patents

Abstract

PURPOSE:To realize a more natural and smooth interaction by generating a goal for expressing knowledge which a user can obtain by speech from the next system, based on a data structure for expressing a problem solving process. CONSTITUTION:A user goal estimating means 3 estimates a user's goal, that is, a user's target state, based on a semantic expression generated by a semantic expression generating means 2. Also, a speech goal generating means 9 generates a speech goal, that is, a goal for expressing knowledge which a user or a system can obtain by speech from the next system, based on a plan goal structure generated by a plan goal structure generating means 8. Moreover, a speech plan generating means 10 generates a speech plan, that is, a behavior train required for attaining the speech goal, based on the speech goal generated by the speech goal generating means 9. A speech plan executing means 11 generates a natural language sentence transferred to the user by executing its speech plan.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dialog processing control device which is used in a dialog processing system for answering questions, searching information, and the like, and which supports user's problem solving in a dialog form with a computer.

[0002]

2. Description of the Related Art FIG. 24 shows, for example, JP-A-63-1536.
The block diagram of the conventional interactive processing control apparatus described in Japanese Patent No. 31 is shown. In the figure, reference numeral 241 is a system for providing information according to a user's request. 242 is a natural language interpreting means for interpreting the meaning of the input sentence from the user, 243 is a command generating means for generating a command to the information providing system 241 according to the meaning of the input sentence by the natural language interpreting means 242, and 244 is for dialogue. The topic management means 245 for managing topics is a gazing object storage means for storing the object currently watched by the user and its transition. In this interactive processing control device, the user's point of gaze is captured by extracting the object gazed by the user, storing it in the gaze object storage means 245, and referring to it when analyzing the subsequent input sentence. I try to have a smooth dialogue.

[0003]

However, in the conventional dialogue processing control device, since the output from the information providing system is presented to the user as it is, when solving the problem of the user interactively. In addition, there is a problem that it is not possible to generate a natural language sentence corresponding to a situation that changes as the dialogue progresses.

The present invention has been made in order to solve such a problem, and provides a goal expressing a state which is a goal of a user and a plan expressing an action sequence necessary for achieving the goal. By asking for information necessary to solve the problem of the system by considering information about such intention, a description that conveys the action to be performed by the user when solving the problem and the reason why the action is necessary, The present invention provides a dialogue processing control device for generating a confirmation sentence for confirming the state of a user, which the system infers, to generate a response according to a situation that changes as the dialogue progresses.

[0005]

A dialogue processing control apparatus according to the present invention is a user goal estimating means for estimating a goal of a user, that is, a state targeted by the user, based on a semantic expression of a natural language sentence, and an estimated user. System goal generating means for generating the system goals necessary to support the achievement of the system goal, and plan goal structure generating means for generating the data structure expressing the action sequence to achieve the generated system goal , A plan goal structure generating means for generating a plan goal structure expressing an action necessary to achieve the generated system goal, and a user or a system obtains by a utterance from the next system based on the generated plan goal structure Utterance goal generation means for generating goals that express knowledge that can be Equipped with an utterance plan generating means for generating the action sequence necessary for the composition, an utterance plan executing means for generating a natural language sentence to be transmitted to the user, and a next utterance predicting means for predicting a Japanese sentence next input by the user It is a thing.

[0006]

In the dialogue processing control device according to the present invention, by adding the plan goal structure generating means and the utterance goal generating means, the question sentence, the explanation sentence, and the confirmation sentence can be correctly generated according to the situation that changes with the progress of the dialogue. be able to.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. 1 is an overall configuration diagram of an embodiment of a dialogue processing control device according to the present invention. In FIG. 1, reference numeral 1 is a natural language sentence storage means for storing a character string input using a keyboard or the like, 2 is a semantic expression generation means for generating a semantic expression of the natural language sentence input by the natural language text storage means 1,
Reference numeral 3 is a user goal estimating means for estimating a goal of the user, that is, a state targeted by the user, based on the meaning expression generated by the meaning expression generating means 2.

Reference numeral 4 is a belief base for storing knowledge about users of the system and goals of the system. Reference numeral 5 is a meaning expression and user goals based on the meaning expressions and user goals obtained by the meaning expression generation means 2 and the user goal estimation means 3. Is a belief base updating means for updating the system, and 6 is a system goal generating means for generating a system goal necessary for supporting the achievement of the goal of the user based on the belief base updated by the belief base updating means 5.

Reference numeral 7 is a problem area knowledge base for storing knowledge about a problem area to be solved by dialogue, and 8 is knowledge of the problem area knowledge base 7, and achieves the system goal generated by the system goal generating means 6. It is a plan goal structure generating means for generating a plan goal structure which is a data structure expressing a plan which is a sequence of actions to be performed and knowledge premised on executing the plan.

Reference numeral 9 denotes an utterance goal generating means for generating an utterance goal based on the plan goal structure generated by the plan goal structure generating means 8, that is, a goal expressing the knowledge that can be acquired by the user or the system by the utterance from the next system. is there. Reference numeral 10 is an utterance plan generation means for generating an utterance plan, that is, an action sequence necessary for achieving the utterance goal, based on the utterance goal generated by the utterance goal generation means 9.

Reference numeral 11 is an utterance plan executing means for executing the utterance plan generated by the utterance plan generating means 10 to generate a natural language sentence to be transmitted to the user.
Reference numeral 12 is a natural language sentence display means for displaying the natural language sentence generated by the utterance plan executing means 11 on the display, and 13 is a Japanese sentence which the user next inputs based on the utterance plan generated by the utterance plan generating means 10. Is a means for predicting the next utterance.

The operation will be described below with reference to FIGS. 2 to 23. FIG. 2 shows an example of interaction between the user and the system in the travel guide system which is an embodiment of the present invention. 21,
Reference numerals 23, 25, and 27 denote inputs from the user, and 22, 2 and
4, 26 and 28 indicate responses from the system.

FIG. 3 shows a third example of the result generated by the semantic expression generating means 2.
1 represents the meaning of the first user input “dropped traveler's check” 21 in the dialog example.

FIG. 4 shows an estimation result example 41 of the user goal estimating means 3, which expresses that "the user has a traveler's check as a goal", and the state of the user "the user drops the traveler's check". And the goal of the user is inferred using a rule expressed by the pair "the goal of the user is having a traveler's check".

FIG. 5 shows a generation result example 51 of the system goal generating means 6 and expresses a goal that the system is going to solve, that is, "the user has a traveler's check as the goal of the system".

FIG. 6 shows a plan goal structure example 601 generated by the plan goal structure generating means 8 and shows a plan which is a sequence of actions made to achieve the system goal and knowledge premised on executing the plan. Express. Specifically, the data 601 is expressed in terms of 9 arguments, and their meanings are as follows. First argument 602
Indicates the name of the plan selected by the system,
It expresses that the system has selected the plan that "the user performs the reissue procedure of the traveler's check at the bank".

The second argument 603 indicates a plan desire condition, that is, a condition necessary for prompting the user to execute the plan shown in the first argument 602. “In order for the user to have the traveler's check, the traveler's check is required. The user knows that it is useful to perform the check reissuance procedure. "

The third argument 604 indicates a condition for applying the plan, that is, a condition that must be satisfied in advance when the plan shown in the first argument 602 is adopted. In order to adopt the plan to do so, the condition that the user has a purchase reservation for traveler's checks must be satisfied. ”

A fourth argument 605 indicates a precondition of the plan, that is, a condition that must be satisfied before the plan shown in the first argument 602 is put into execution, and "a user performs a traveler's check reissue procedure. Before that, the user must be in the bank, and in order for the user to be in the bank, the plan is for the user to go to the bank. "

The fifth argument 606 is the plan schema condition, that is, the plan indicated by the first argument 602 is the ninth argument 6
10 shows a condition necessary for having the goal shown by 10 as an effect, in order for the user to have a traveler's check as a result of the user's reissue procedure for the traveler's check appearing. You need to have a lost certificate of check, and to have a lost certificate of traveler's check, you need to apply for a plan to apply for a lost certificate at the police station. "

The sixth argument 607 indicates an action sequence executed when the plan is decomposed, that is, the plan shown in the first argument 602 is realized. "A user does not perform a traveler's check reissue procedure, It is expressed as "a user goes to the window", "it makes an application for reissue of traveler's check", and "receives traveler's check".

The seventh argument 608 indicates a deletion proposition, that is, a proposition expressing a state that is not satisfied after the execution of the plan shown in the first argument 602. In the case of this example, "[]", that is, there is no particular Express that.

The eighth argument 609 indicates an additional proposition, that is, a proposition expressing a state that is established after the execution of the plan shown in the first argument 602. “When a user applies for a traveler's check, he / she has a traveler's check. The result is obtained. ”

The ninth argument 610 indicates the goal of the plan, that is, the goal to be achieved by adopting the plan shown in the first argument 602. The reason for adopting the plan is to achieve the goal that the user has a traveler's check. ”

FIG. 7 is an example knowledge 71 about a problem area stored in the problem area knowledge base. Specifically, it is expressed in terms of 7 arguments, and each meaning is as follows. The first argument 72 indicates an action name, and expresses that the knowledge 71 is knowledge about the action “the user goes to the bank”. The second argument 73 indicates the application condition of the action, and expresses “[], that is, there is no particular condition that must be satisfied before the user executes the action of going to the bank”.

The third argument 74 indicates the precondition of the action, and expresses that "in order for the user to go to the bank, the condition that the user knows the location of the bank must be satisfied". .

The fourth argument 75 indicates the schema condition of the action, and the condition necessary for "[], that is, the condition that the user is in the bank to appear as a result of executing the action of the user going to the bank, is "There is nothing special". The fifth argument 76 indicates the decomposition of the action, and expresses “[], that is, the action of the user going to the bank is an action that cannot be further divided in this area plan”.

The sixth argument 77 indicates a deletion proposition of the action, and expresses that "as a result of the user going to the bank, the user is no longer at the place where he was before going to the bank". The seventh argument 78 indicates an additional proposition of the action, and expresses that “the user is in the bank as a result of the user going to the bank”.

FIG. 8 is a flow chart showing the processing contents of the plan goal structure generating means 8. When the process of the plan goal structure generating means 8 is started, in step 801, the system goal generated by the system goal generating means 6 is received as an input.

In step 802, step 80
The problem domain knowledge base is searched for knowledge about an action having the system goal input by 1 as an additional proposition. In step 803, it is determined whether or not the search in step 802 was successful.
If the step 804 fails, the process ends. In step 804, the knowledge about the act obtained in step 802 is converted into a plan goal structure.

More specifically, the arguments indicating application conditions, preconditions, schema conditions, decompositions, deletion propositions, and additional propositions in the plan goal structure correspond to the knowledge in the action obtained from the problem domain knowledge base. Copy the conditions, [] for the argument indicating the desire condition, and step 8 for the argument indicating the goal.
Substitute the system goal obtained by 01.

In step 805, step 80
It is determined whether or not the application condition of the plan goal structure obtained in step 4 is satisfied. If the condition is satisfied, step 806 is performed. If not satisfied, step 802 is performed.
Branch to. In Step 806, Step 80
It is determined whether or not the preconditions of the plan goal structure obtained in step 4 are satisfied. If they are satisfied, step 810 is performed. If they are not satisfied, step 807 is determined.
Branch to.

In step 807, step 80
The process of the plan goal structure generating means is recursively called with the system condition as the precondition for determining whether or not the condition is satisfied. In Step 808, Step 80
It is determined whether or not the recursive call of 7 is successful. If it is successful, the process branches to step 809, and if it is not successful, the process branches to step 802. In step 809, the plan goal structure generated in step 808 is substituted into the precondition of the plan goal structure obtained in step 804.

In step 810, step 80
It is determined whether or not the schema condition of the plan goal structure obtained in step 4 is satisfied. If it is satisfied, the process branches to step 811, and if not, the process branches to step 812. In Step 812, Step 811
It is determined whether or not the recursive call of is successful. If it is successful, the process branches to step 813, and if it is not successful, the process branches to step 814.

In step 813, step 81
The plan goal structure obtained in 1 is substituted into the argument indicating the schema condition of the plan goal structure obtained in step 803. In step 814, the desire condition of the plan goal structure is added. More specifically, the proposition indicating that the system plan indicated by the plan name of the plan goal structure is useful for the goal is substituted for the argument indicating the desire condition of the plan goal structure.

Next, in the case where the belief base 4 contains only the proposition "to have (user, refrain from purchasing traveler's checks)", that is, "the user has refrain from purchasing traveler's checks", as a proposition expressing the situation, An example of the processing flow of the system generation means when the system goal shown at 51 in FIG. 5 is input will be described with reference to FIGS. 9 to 14.

In step 801, the system goal 51 is input as a system goal. Next, in the process of step 802, the plan 91 shown in FIG. 9 is searched from the problem area knowledge base, and since the search is successful, the process branches to step 804.

In step 804, the plan goal structure 101 shown in FIG. 10 is generated, and in step 805, the application condition “have (user, user,
"Travelers check purchase refusal") 102 is established in the belief base 4, so the process branches to step 806.

In step 806, the precondition “Iru (user, bank)” 103 of the plan goal structure 101 is not satisfied in the belief base 4, so that step 807 is performed.
Branch to. In step 807, the plan goal structure generating means 8 is called by using "Iru (user, bank)" as the system goal, and the plan goal structure 111 shown in FIG. 11 is generated, and the recursive call is successful. Branch off.

In step 809, the plan goal structure 121 shown in FIG. 12 is generated. Next, Step 81
In 0, the schema condition “have (user, lost certificate)” 104 of the plan goal structure 101 is not satisfied in the belief base 4, so the process branches to step 811.

In step 811, the plan goal structure 131 shown in FIG. 13 is generated, and the processing in step 812 branches to step 813 to generate the data 141 showing the plan goal structure shown in FIG. Finally,
In step 814, the plan goal structure 601 is generated.

FIG. 15 shows an utterance goal example 151 generated by the utterance goal generating means 9 from the plan goal structure example 181 and expresses a proposition expressing the knowledge of the system or the user trying to achieve by the utterance of the next system. . Specifically, it indicates that “the user knows the location of the bank near the user's whereabouts” is the goal that the utterance of the next system is trying to achieve.

FIG. 16 is a flowchart showing the operation of the utterance goal generating means 9. When the process of the utterance goal generating means 9 is activated, in step 161, the plan goal structure generated by the system plan generating means 8 is received as an input. In step 162, a proposition regarding the knowledge of the user or the system whose establishment in the belief base 4 cannot be confirmed is extracted from the plan goal structure, and the set of extracted propositions is output.

In step 163, the evaluation value of each extracted utterance goal candidate is calculated by using the evaluation means shown in FIG. In Step 164, Step 1
The utterance goal candidate having the maximum evaluation value is extracted from the utterance goal candidates calculated at 63, and the utterance goal is output from the utterance goal generating means 9.

FIG. 17 is a flow chart showing the processing contents of step 163. In step 171, the set of utterance goal candidates obtained as the output of step 162 is input. In Step 172, Step 17
It is judged whether there is a set of utterance goal candidates obtained in 1 for which the evaluation value has not been calculated yet. To branch to step 173.

In step 173, step 171
From the set of utterance goal candidates obtained in step 1, select one utterance goal candidate for which an evaluation value has not been calculated. Step 1
At 74, a value V1 for evaluating the depth of the relationship with the utterance goal corresponding to the previous utterance is obtained. Specifically, the plan-goal structure generated by the plan-goal structure generating means 8 is grasped as a graph as shown by 182 in FIG. 18, and obtained in step 173 from the node expressing the utterance goal corresponding to the previous utterance. The reciprocal of the number of arcs on the graph required to reach the node expressing the utterance goal is calculated, and the value is set to V1.

In step 175, step 173
A value V2 that evaluates the magnitude of the effect that the achievement of the utterance goal obtained by is given to the achievement of the user's upper goal is calculated. Specifically, when a kind of condition is given, a function that returns a value indicating the importance of the condition is used, and the value is obtained by acting on the function of the kind of condition in the plan goal structure of the utterance goal candidate, The value is set to V2. Conditions in the plan goal structure are assumed to have a high degree of importance in the order of application conditions, schema conditions, preconditions, and desire conditions. For example, as a function that associates a condition type with a value,
It is assumed that 10, 7, 5, and 3 are associated in the order of the above-mentioned conditions.

In step 176, step 173
A value V3 that evaluates how appropriate the utterance goal obtained by is viewed from the problem solving ability of the user is obtained. Specifically, using the function shown in FIG. 19 that returns a value indicating the importance of the utterance goal from the user's comprehension and the type of condition, the condition in the plan goal structure of the utterance goal candidate is applied to the function. The value is obtained by setting the value as V3.

In step 177, step 173
A value V4 for evaluating whether or not the utterance goal obtained by is satisfied is obtained. For example, as the value of V4, 0 is included if the proposition expressing the utterance goal candidate is included in the belief base 4, 1 is included if the proposition expressing the utterance goal candidate is not included, and a proposition denying the proposition expressing the utterance goal is included. In case of 2

In step 178, step 17
Evaluation value V obtained from 4 to step 177
The product of the values of 1, V2, V3, and V4 is taken, and the value is set as the evaluation value of the utterance goal obtained in step 173.

FIG. 18 shows the correspondence between the expression 181 using the terms of the plan goal structure and the expression 182 using the graph.

FIG. 19 is an example of a function that returns a value indicating the degree of importance of a condition based on the user's understanding amount and the type of condition. The user comprehension is shown in (,) on the left side of → in the line of the figure. The right side of → shows a value that evaluates the importance of the utterance goal viewed from the user's problem solving ability. When the user's comprehension is large, the desire condition is emphasized. When the user's comprehension is small, the desire condition is not emphasized. Furthermore, the application condition is emphasized regardless of the user's comprehension. There is.

Next, when the belief base 4 stores a proposition that the user's comprehension is small, the utterance "where are you" 21 in FIG. 2 is generated when the system generates the utterance. An example of the processing flow of the goal generating means will be described with reference to FIGS. 20 and 21.

In step 161, the plan goal structure 181 shown in FIG. 18 is input, and in step 162, the set of utterance goal candidates shown in FIG. 20 is obtained.
In step 163, the sequence of pairs of utterance goal candidates and evaluation values shown in FIG. 21 is obtained.

This evaluation value is calculated according to the flowchart shown in FIG.
If the value is 1, the value of V1 is the same as the previous utterance "Do you have a lost certificate of traveler's check?" 24, "Yes"
The node "true (has (us (us
er, lost certificate)) ”183 corresponding to the utterance goal candidate 151 on the graph 182“ know-ref (user, X,
Location, bank, []) ”The number of arcs to reach 184 is 1/3 because it is 3 and the value of V2 is node 1
Since 83 is a prerequisite, the value becomes 5, and the value of V3 is 2 because the understanding of the user is small and it is a prerequisite.
Therefore, the value of V4 becomes 1 because the proposition indicating that this goal is satisfied is not stored in the belief system.

As a result, in step 178, the evaluation value of the utterance goal candidate 151 becomes 3.33. The evaluation values of other goal candidates are calculated in the same manner. Step 164
, The utterance goal candidate “know-r
ef (user, X, place, bank, []) ”211 is extracted, and the output of the utterance goal generation means 9 is set as the utterance goal 151.

FIG. 20 shows an example of a set of utterance goal candidates obtained as the output of step 162.
"Users will find it useful to have travelers checks at the bank to apply for reissue of travelers checks.""Users will know that going to banks will help to have travelers checks.""User knows where bank is close to where he / she is.""User knows it is useful to have a traveler's check to issue a lost certificate.""Go to police station has a traveler's check." It is expressed that the user knows that it is useful for "the user knows the location of the police station near the user's whereabouts".

FIG. 22 shows an utterance plan example 221 generated by the utterance plan generating means 10 and expresses an action sequence necessary for achieving the utterance goal generated by the utterance goal generating means 9. Specifically, the utterance goal shown in “User knows bank location” 151 is “System asks user whereabouts user” 222,
"User communicates location to system" 223, "System retrieves location of bank closest to user" 22
4, "The system communicates the bank's location to the user" 22
It shows that it is achieved by four actions of 5. Further, the utterance plan executing means 11 executes the “system asks whereabouts the user” 222 to generate the utterance 26.

FIG. 23 shows the action 231 of the user predicted by the next utterance prediction means 13, "the user tells the system where the user is".
Express that. This action is based on the sequence of actions remaining after removing the system's performed actions from the speech plan.
Obtained by the user selecting the next action to perform.

In the above example, a dialogue example in which the goal of the user does not conflict with the goal of the system has been shown. However, when the goal of the user and the goal of the system do not conflict, the difference between the goal of the user and the goal of the system is shown. It is possible to realize a dialogue for dealing with a user's misunderstanding by adding the processing for checking.

[0061]

As described above, according to the present invention, the problem obtained by the cooperation between the system and the user is obtained by using the knowledge obtained from the utterances made before and the knowledge about the problem area to be solved by the dialogue. By generating goals expressing the knowledge that can be acquired by the user or the system by the utterance from the next system based on the data structure expressing the solution process, question sentences and explanations according to the situation that changes as the dialogue progresses Sentences and confirmation sentences can be generated, and more natural and smooth dialogue can be realized.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of an embodiment of a dialogue processing control device according to the present invention.

FIG. 2 is an explanatory diagram showing an example of dialogue in an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing an example of a result generated by a semantic expression generating means.

FIG. 4 is an explanatory diagram showing an example of an estimation result of a user goal estimation means.

FIG. 5 is an explanatory diagram showing an example of a generation result of a system goal generating means.

FIG. 6 is an explanatory diagram showing an example of a generation result of a plan goal structure generating means.

FIG. 7 is an explanatory diagram showing an example of knowledge stored in a problem area knowledge base.

FIG. 8 is a flowchart showing the operation of the plan goal structure generating means.

FIG. 9 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 10 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 11 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 12 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 13 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 14 is an explanatory diagram showing an example of data generated during the operation of the plan goal structure generating means.

FIG. 15 is an explanatory diagram showing an example of an utterance goal generated by an utterance goal generating means.

FIG. 16 is a flowchart showing the operation of the utterance goal generating means.

FIG. 17 is a flowchart showing a method of calculating an evaluation value of an utterance goal that appears during the operation of the utterance goal generating means.

FIG. 18 is an explanatory diagram showing a method of expressing a plan goal structure.

FIG. 19 is an explanatory diagram showing functions used when calculating evaluation values of utterance goal candidates.

FIG. 20 is an explanatory diagram showing a set of utterance goal candidates.

FIG. 21 is an explanatory diagram showing pairs of utterance goal candidates and their evaluation values.

FIG. 22 is an example of an utterance plan generated by an utterance plan generating means.

FIG. 23 is an explanatory diagram showing a user's action predicted by the next utterance prediction unit.

FIG. 24 is an overall configuration diagram of a conventional dialogue processing control device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Natural language sentence storage means 2 Semantic expression generation means 3 User goal estimation means 4 Belief base 5 Belief base update means 6 System goal generation means 7 Problem domain knowledge base 8 Plan goal structure generation means 9 Utterance goal generation means 10 Utterance plan generation means 11 utterance plan executing means 12 natural language sentence displaying means 13 next utterance predicting means 241 information providing system 242 natural language interpreting means 243 command generating means 244 topic managing means 245 gazing object storage means

Claims (1)

[Claims] 1. A natural language sentence storage means for storing an input character string in an interactive processing system for interpreting the content of an input sentence, solving a problem, and generating a response sentence.
A semantic expression generating means for generating a semantic expression of the input natural language sentence, a user goal estimating means for estimating a target state (goal) of the user based on the generated semantic expression,
A belief base that stores knowledge about the users of the system and the goals of the system itself, a belief base updating means that updates the semantic representation and the user goals based on the generated semantic representation and the estimated user goals, and a belief base based on the updated belief base. System goal generating means for generating system goals necessary to support the achievement of user goals, a problem domain knowledge base for storing knowledge necessary for a problem domain to be solved by dialogue, knowledge of this problem domain knowledge base Plan goal structure generation means for generating a plan goal structure, which is a data structure expressing a plan that is a sequence of actions performed to achieve the generated system goal and knowledge that is assumed when the plan is executed, by using , Based on the generated plan goal structure, the utterance from the next system (utterance goal) Utterance goal generating means for generating a goal expressing knowledge that can be acquired by a user or a system, utterance plan generating means for generating an utterance plan based on the generated utterance goal, that is, an action sequence necessary for achieving the utterance goal, utterance A speech plan execution means for generating a natural language sentence to be transmitted to the user by executing the plan, a natural language sentence display means for displaying the generated natural language sentence, and a speech plan generated by the speech plan execution means. A dialogue processing control device comprising a next speech prediction means for predicting a Japanese sentence to be input next by the user.