Dialogue FAQ Generation Using Emotion and Sentence Variance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing operator support systems in call centers lack an automated method for generating frequently asked questions (FAQs) that consider the dependency of answer sentences on question sentences and questioner emotions, as described in JP 2020-115244 A.
Innovation Solution
A dialogue support system that automatically generates question/answer pairs by considering the similarity between answer and question sentence vectors and emotion vectors, distinguishing between sentence-dependent and emotion-dependent pairs, and stores them in a database for operator assistance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If an operator support system uses existing FAQ databases and emotion analysis to provide support, then operators receive assistance during customer interactions, but the system lacks automated generation of context-dependent FAQs that consider both question sentences and questioner emotions
Solution Approach 1:
The system segments the FAQ generation task into multiple processing stages: collecting dialogue data, extracting question/answer pairs, analyzing emotion vectors, calculating variances, and generating categorized FAQs. This segmentation enables automated processing while preserving context dependency information through structured analysis at each stage.
Solution Approach 2:
The system changes parameters by introducing variance analysis of emotion vectors and question sentence vectors as key parameters. By calculating and comparing variances, the system automatically determines whether to generate emotion-dependent or sentence-dependent FAQs, enabling context-aware automated generation without manual intervention.
2Measurement precision
If the system generates FAQs considering both question sentences and questioner emotions, then the accuracy and relevance of operator support improves, but the complexity of the dialogue support system increases
Solution Approach 1:
The system implements dynamic FAQ generation by adjusting the generation approach based on variance calculations. When emotion vector variance is low, the system generates emotion-dependent FAQs; when question sentence vector variance is low, it generates sentence-dependent FAQs. This dynamic adaptation improves accuracy while managing complexity through conditional logic rather than fixed complex architecture.
Solution Approach 2:
The system introduces variance analysis as an intermediary mechanism between input dialogue data and output FAQs. By using variance calculations as an intermediate processing step, the system simplifies the complexity of directly mapping complex contextual information to appropriate FAQ types, enabling accurate generation through a structured intermediate analysis layer.
3Quantity of substance
If the system processes and analyzes all dialogue data to generate comprehensive FAQs, then the coverage and completeness of operator support improves, but the processing time and computational resources increase
Solution Approach 1:
The system applies partial action by focusing processing on variance analysis of vectors rather than analyzing every aspect of dialogue data in detail. By calculating variances as a simplified metric, the system achieves comprehensive FAQ coverage without the excessive time cost of exhaustive analysis, processing only the essential statistical properties needed for categorization.
Solution Approach 2:
The system changes the processing approach by transforming detailed dialogue analysis into variance calculations of aggregated vectors. This parameter transformation reduces computational complexity from analyzing individual dialogue elements to calculating statistical measures, enabling comprehensive FAQ generation with reduced processing time and resources.
Data Source
AI summary
A dialogue support system holds question/answer pair candidates including a question sentence vector, an emotion vector indicating a question emotion, and an answer sentence vector, outputs a question/answer pair based on question/answer pair candidates included in a first group as a sentence dependence question/answer pair in a case where it is determined that a variance of emotion vectors included in a first group based on similarity between answer sentence vectors and similarity between the question sentence vectors is large, and outputs a question/answer pair based on question/answer pair candidates included in a second group as an emotion dependence question/answer pair in a case where it is determined that a variance of question sentence vectors included in a second group based on similarity between the answer sentence vectors and similarity between the emotion vectors is large.