Conversation Evaluation Apparatus Using Overlap and Silence Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing conversation evaluation technologies cannot objectively assess the impression of the first speaker independently from the second speaker, leading to unfair evaluations, especially when the second speaker cuts in while the first speaker is speaking.
Innovation Solution
A conversation evaluation apparatus and method that estimates the starting and ending times of each main speaker's utterance, identifies the timing of switches between main speakers, and evaluates the conversation state based on dialogue information including overlapping and silent segments, using a detection model trained on dialogue data to provide objective evaluations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the evaluation method uses only the overlapping segment length between two speakers, then the evaluation process is simple, but the evaluation result becomes unfair and inaccurate when one speaker cuts in while the other is speaking
Solution Approach 1:
The patent segments the conversation evaluation into multiple independent components: overlapping segment analysis, silent segment analysis, and turn-switch timing analysis. Each segment is evaluated separately and then integrated to produce a comprehensive assessment, allowing fair evaluation of each speaker's contribution without being biased by interruptions
Solution Approach 2:
The patent adds new evaluation dimensions beyond just overlapping segment length. It introduces silent segment length analysis and turn-switch timing analysis as additional dimensions, transforming the evaluation from a single-dimensional metric to a multi-dimensional assessment that captures the full dynamics of speaker interactions
2Ease of operation
If the evaluation considers only the second speaker's position, then the evaluation method is straightforward, but it cannot objectively evaluate the first speaker's impression independently
Solution Approach 1:
The patent inverts the traditional evaluation perspective by analyzing the conversation from both speakers' viewpoints rather than just the second speaker's position. It evaluates how the first speaker's utterance is affected by the second speaker's interruption, providing a balanced assessment that captures both perspectives equally
3Measurement precision
If the evaluation uses multiple dialogue information parameters, then the evaluation accuracy improves, but the analysis complexity and processing time increase
Solution Approach 1:
The patent extracts and focuses on the three most critical dialogue information parameters: overlapping segment length, silent segment length, and turn-switch timing. By selecting only these essential parameters rather than analyzing all possible conversation features, it achieves high evaluation accuracy while keeping processing complexity manageable
Data Source
AI summary
According to one embodiment, a non-transitory computer readable medium includes computer executable instructions. The instructions, when executed by a processor, cause the processor to perform a method. The method estimates a starting time and an ending time of an utterance of each main speaker relating to a conversation. The method identifies a timing of a switch between the main speakers. The method evaluates a state of the conversation based on dialogue information before and after the identified timing of the switch. The dialogue information includes at least one of a length of an overlapping segment or a length of a silent segment, and includes a length of an utterance segment.