JPH0344690A - Conversation manner education system - Google Patents

Abstract

PURPOSE:To recognize and correct an error in kind which can not be corrected by a conventional system by using grammatical knowledge regarding terms of honorific language using knowledge for the terms of honorific language state information on a conversation, and a conversation pattern. CONSTITUTION:A recognition part 4 converts an input voice into a phoneme string, a language processing part 5 converts the string into character information, and a conversation understanding part 6 extracts the meaning of a sentence from the character information. A interaction control part 7 extracts and holds necessary state information to generate a response. A honorific language generation part 8 generates a conversation with correct expression by using the meaning contents of the conversation. An evaluation part 9 compares and analyzes the character information and the converted sentence with correct expression to evaluate the expression and contents of the conversation. A total evaluation part 10 performs total evaluation. An output part 11 outputs indications to a user, a response to the conversation, and evaluation in the form of voice or character information. The conversation understanding part 6 uses the state knowledge 14, the respect term generation part 8 uses the honorific language grammatical knowledge 12, and usage knowledge 13, and the interaction control part 7 and evaluation part 9 uses the data base of interaction patterns 15 while the interaction control part 7 uses image data 16 according to an interaction advance. Consequently, the user easily learns the conversion based principally upon honorific language close to a real conversation.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention is designed to improve the use of language and desirable conversation in accordance with social conventions, which are necessary in public dialogue situations such as answering the telephone, customer service, interviews, and self-introductions. This invention relates to a conversation etiquette education system using a computer device for the purpose of supporting learning the development of conversation.

[Prior Art] An example of something similar to the intended use of the present invention is an English conversation education system that allows users to learn English conversation while freely conversing with the system. This English conversation education system has a function that points out grammatical errors and spelling errors in the user's utterances while having a free conversation in English with the user, and suggests appropriate phrases for the utterance that suit the situation. be. In order to realize such functions, this English conversation education system provides knowledge such as a word dictionary, a grammar dictionary, and specific conversation patterns. The user inputs the conversation using the keyboard, the input text is displayed on the display, and the system's response is output on the display and by voice.

[Problem to be solved by the invention] In conventional language education systems, correction of errors in input sentences is done by
This is supported by word dictionaries, grammar dictionaries, and specific conversation patterns. Errors in conversation manners, which are the target of this invention, are not only grammatical, but also errors due to errors in the use of honorific language, inconsistency with the conversation development pattern, etc., so conventional error correction functions cannot handle them. Can not.

[Means for solving the problem] As a means of recognizing and correcting errors of a type that cannot be handled with conventional technology, we use grammatical knowledge about honorific language, knowledge of how to use honorific language, conversation situation information, and conversation patterns.

[Operation] After the conversational speech input by the user is recognized and understood as an input sentence by the speech recognition unit, language processing unit, and speech understanding unit, the evaluation unit evaluates it as honorific grammar knowledge, honorific usage knowledge, and conversation situation information. Errors in the expression of sentences are recognized using . Furthermore, it is compared with the conversation development pattern to determine whether the input sentence is appropriate for the conversation development. The dialogue management unit generates a corresponding system response from these processing results and conversation buttons.

This repetition creates a conversation between the system and the user. When the conversation ends, the education strategy department outputs an overall evaluation of the conversation and advice regarding the use of words, etc., based on the user's saved utterances.

[Embodiment] FIG. 1 is a block diagram showing an embodiment of a device for configuring the system of the present invention. 2 is an audio input/output device, 2 is a display device, and 3 is a processing device for processing all information.

The user transmits audio to the processing device 3 using an audio input/output device connected to the processing device via a telephone line or directly.
to use human power.

The processing bag@3 recognizes the user's voice converted into an electrical signal, analyzes and understands the signal, generates a corresponding response sentence, sends it to the voice input/output device, outputs the voice, and communicates with the user. Proceed with the conversation. We also provide education on conversation etiquette. Furthermore, by outputting text and image information on the display 2, it helps the user's understanding.

FIG. 2 is a block diagram showing the configuration of each processing section within the processing device. In FIG. 2, reference numeral 4 denotes a recognition unit, which converts input speech into a phoneme string using a speech recognition R function. The language processing unit 5 converts the phoneme string into character information. The speech understanding unit 6 is
Extract the meaning of a sentence from character information.

The utterance understanding unit 7 receives the content of the user's utterance, extracts and stores necessary information as situation information, and further generates a response as a system. The honorific expression generation unit 8 receives the meaning of the utterance and generates an utterance with a correct expression.

The evaluation unit 9 compares and analyzes the user's utterance converted into text information and the fe version of the utterance content into a sentence using correct expressions, and evaluates the expression and content of the user's utterance. do.

The comprehensive evaluation unit IO analyzes the evaluation of the user's utterances and generates a comprehensive evaluation. The output unit 11 outputs image information to the display, and outputs instructions to the system user, system responses and evaluations to the user's utterances as voice and text information.

The situational knowledge database 14 is used by the speech understanding unit 6, the honorific language grammar knowledge database 12 and the honorific language usage knowledge database 13 are used by the honorific language generation unit 8, and the dialogue pattern database 15 is used by the dialogue management unit 7 and evaluation unit 9. used respectively.

The image data is used by the dialogue management section 7 according to the progress of the dialogue.

FIG. 3 is a flowchart showing the processing procedure within the processing device, and the reference numbers of the corresponding processing units in FIG. 2, which are responsible for each process, are shown in parentheses. Using FIG. 3, the processing of the system will be explained in accordance with the processing order.

The processing series in this system can be broadly divided into two types: a dialogue series and an evaluation series.

First, to explain along the dialog sequence, input speech is recognized by the recognition section 4 and converted into character information by the language processing section 5. This character information is output from the output unit 11 as characters. Furthermore, this character information is understood by the speech understanding unit 6,
The understood content is analyzed by a dialogue pattern in the dialogue management section 7, and a response content is generated. The content of this response is converted into a correctly expressed sentence by the honorific language generation section 8, and outputted from the output section 11 as audio and text information. At the same time, image data corresponding to the situation, obtained by the dialog management section 7 by searching for image data, is outputted as an image from the output section 11.

Next, to explain along the evaluation series, the character string containing the user's utterance converted into character information in the language processing unit 5, and the correct character string that is understood in the utterance understanding unit 6 and converted in the honorific expression generation unit 8. The user's utterances with expressions are compared and evaluated in the evaluation section 9.

This comparison and evaluation is based on the Keigo Usage Knowledge Database 1, which describes the human relationships necessary for using honorifics in each conversation situation.
This is done using a honorific grammar knowledge database 13 that describes the respectful form, humble form, and polite form of the 01 word, which can be rewritten, for each word. FIG. 8 illustrates a part of the honorific language usage knowledge database 12 and the honorific language grammar knowledge database 13. Specifically, this comparison and evaluation is based on the user's rI! t
This is done by finding erroneous parts such as non-use or misuse of 55 words, mixing up honorific words and humble words, storing them, and tallying error scores for each type of error.

Furthermore, the content of the utterance is also evaluated by referring to the dialogue pattern. This evaluation uses dialogue pattern 15, which describes the desired dialogue flow for each conversation situation as a series of dialogue pairs between the system and the user, to ensure that the contents of the actual user's utterances do not deviate from this dialogue pattern 15. This is done by checking the FIG. 9 illustrates a part of the dialogue pattern 15. Specifically, when the order and content of the user's utterances deviate from this dialogue pattern, the system saves the utterances that deviated from the dialogue pattern and the desired original utterance content, and then calculates the error distance score, which is the degree to which the utterances deviate. calculate.

When the dialogue between the system and the user is completed, the comprehensive evaluation section 10
In the process, the evaluation results of the user's utterances are comprehensively evaluated, and each evaluation is outputted to the output unit 11 as text information and audio information.
is output from. The evaluation output from the comprehensive evaluation section 10 includes errors in the expression of the user's utterances and correct answers, indications of undesirable utterance contents, examples of desirable correct answers, and further evaluations based on the evaluation results received from the evaluation section 9. From Part 9, there is a comprehensive evaluation of the user's utterances derived from the error type score of the utterance expression and the error distance of the utterance content, and hints for correct conversation. Regarding the comprehensive evaluation and hints for correct conversation, the output content is prepared in advance in the form of a database within the comprehensive evaluation department, and appropriate content is extracted using the error type score and error distance as keys. .

FIG. 4 shows a flowchart showing the operation of the system.

FIG. 5 is a flowchart illustrating in detail the operation of the conversation mode shown in FIG. FIG. 6 is a flowchart illustrating in detail the operation of the evaluation mode shown in FIG. FIG. 7 is a flowchart illustrating in detail the operation of the learning mode shown in FIG.

The output of the system to the user will be explained based on the flowcharts of FIG. 4 and related FIGS. 5-7.

When a practice conversation scene is selected, an explanation of the set situation is outputted to the display 2 as voice information and text information.

In addition, in the conversation mode, the device operates according to the procedure shown in the flowchart in Figure 5, and the set conversation status is output as image information to the display 2, and the conversation between the system and the user can be recorded as voice or text. Output as information.

Furthermore, the system determines whether the expression and content of the user's utterances are close or far from the correct sentence expression and dialogue pattern, and displays the degree of distance between the output of the buzzer sound and the screen of Display 2. Represented by changes in human facial expressions.

In the evaluation mode, the device of the present invention operates according to the flow shown in FIG. 6, and outputs an evaluation regarding the content and expression of the user's conversation using voice and text information, and also outputs an overall evaluation of the user's conversation. is output.

In the learning mode, which is used when the user has not learned enough conversation, the device operates according to the flow shown in Figure 7, and provides hints for correct conversation, examples of correct conversation, and other possible utterances. The content and way of expression are output as audio and text information during conversation learning with the system, which helps in learning correct conversation. Figure 10 shows part of an example of a conversation between the system and the user. FIG. 11 shows a specific operational example of conversation evaluation. Figure 12 shows an example of conversation evaluation. FIG. 13 is an example of the overall evaluation of the conversation, hints, and advice. Figures 10 to 13 are examples of evaluations of user conversations of the system described in Figures 5 to 7.

[Effects in January] The present invention evaluates the user's utterances in specific situations set by the system from two points: the use of language centered on honorific language and the development of conversation topics. It provides appropriate advice, and therefore uses not only general grammatical knowledge of honorific language, but also practical knowledge of honorific language, situational knowledge of the conversation world, and conversation development patterns for each specific conversation situation. By doing this, it is now possible to support users in learning not only the content of conventional conversational sentences, but also the language used, the way the topic flows, and other problems that change depending on the situation. . The user's utterances are understood by voice, and the system outputs a response sentence in a manner appropriate to the conversation situation, allowing the user to quickly and easily converse with the system, making it more realistic. Learning can be done.

[Brief explanation of drawings]

Figure 1 is a block diagram of an embodiment of the present invention, Figure 2 is a functional block diagram of the processing device, Figure 3 is a flowchart showing the processing procedure within the processing device, and Figure 4 shows the operation of the system. Flowcharts: Figure 5 is a flowchart explaining the operation of the conversation mode in Figure 4; Figure 6 is a flowchart explaining the operation of the evaluation mode in Figure 4; Figure 7 is a flowchart explaining the operation of the learning mode in Figure 4. This is a flowchart. Also, Figure 8 shows an example of the contents of the honorific language usage knowledge database and the honorific grammar knowledge database, and Figure 9 shows an example of a dialogue pattern.
The figure shows an example of a part of a conversation between the system and the user, Figure 1 shows a concrete operation example of conversation evaluation, Figure 12 shows an example of conversation evaluation, and Figure 13 shows the overall conversation evaluation, hints, and advice. This is an example.・Voice input/output device・Processing device・Language processing section・Dialogue management section・Evaluation section・Display・Recognition section・Speech understanding section・Keigo generation section・Comprehensive evaluation section II・-Output section 12・・Keigo grammar knowledge 13・・Keigo usage knowledge 14・・Situation knowledge 15・・Dialogue pattern 16・・Image data

Claims (1)

[Claims] A speech recognition unit that recognizes the input user's voice as phonemes and character strings, a language processing unit that generates input sentences from the recognized character strings, a conversation understanding unit that understands input sentences by comparing them to the conversation situation, and honorific language. a large-capacity storage device serving as a database that stores information on grammatical knowledge about, information on conversation situations, knowledge on using honorific language according to conversation situations, and information on conversation patterns for progressing the conversation; honorific grammar knowledge read from the database; Honorific usage knowledge, conversation situation information,
An evaluation unit that evaluates the expression and content of an input conversation sentence using information regarding conversation patterns; a dialogue management unit that generates a response corresponding to the input sentence by referring to the conversation situation and conversation pattern; and a generated response sentence. A comprehensive evaluation section that analyzes the evaluation of each user's conversation sentences, determines an educational policy for learning conversation manners, and outputs learning advice to the user. A conversation etiquette education system comprising: a response voice corresponding to the input voice, and an output unit that outputs character image information to support user learning.