Conversation Evaluation Device Pitch Interval Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques fail to objectively evaluate the non-linguistic characteristics of responses in conversations, specifically the impression given to the conversation partner, by not comparing voice characteristics of questions and responses, and do not account for dynamic changes in prosodic accommodation.
Innovation Solution
A conversation evaluation device and method that compares the representative pitches of questions and responses, adjusting pitch differences on an octave-by-octave basis to ensure a predetermined range, and evaluates the conversation interval to assess the impression given by a response, using a CPU-driven system with voice input and analysis sections to calculate evaluation scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pitch comparison between question and response is performed without octave adjustment, then evaluation precision is improved, but adaptability deteriorates due to inability to handle significant pitch differences between speakers
Solution Approach 1:
The patent applies parameter changes by adjusting pitch values through octave shifting. When the pitch difference between question and response exceeds a predetermined threshold, the system shifts the response pitch by one or more octaves to bring it within the comparison range, enabling accurate evaluation across different speaker pitch ranges while maintaining evaluation precision
Solution Approach 2:
The system dynamically adjusts the pitch comparison process by first calculating the actual pitch difference, then conditionally applying octave shifts based on whether this difference exceeds the threshold. This dynamic adaptation allows the evaluation system to handle both same-pitch and different-pitch speaker scenarios appropriately
2Device complexity
If only linguistic content is analyzed, then device complexity is reduced, but evaluation reliability deteriorates due to inability to assess non-linguistic characteristics
Solution Approach 1:
The patent merges linguistic content analysis with non-linguistic characteristic analysis (pitch and time intervals). By combining these multiple analysis dimensions into a unified evaluation system, the patent achieves comprehensive conversation evaluation that assesses both what is said and how it is said, thereby improving evaluation reliability without excessive complexity increase
3Measurement precision
If pitch interval evaluation is performed without considering consonant intervals, then measurement precision is improved, but evaluation reliability deteriorates due to ignoring musically significant relationships
Solution Approach 1:
The system performs preliminary action by pre-calculating and storing consonant interval relationships (such as perfect fifths, fourths, and octaves) before the actual pitch comparison. During evaluation, it checks whether the measured pitch interval corresponds to these pre-established consonant relationships, thereby incorporating musical significance into the precision measurement without compromising either aspect
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Information related to voice of a question and information related to voice of a response to the question are received. On the basis of the received information, an analysis section (106) acquires a representative pitch of the question (e.g., a pitch of the end of the question), and a representative pitch of the response (e.g., an average pitch of the response). Further, on the basis of comparison between the representative pitch of the question and the representative pitch of the response, an evaluation section (110) evaluates the voice of the response to the question on the basis of how much a difference (interval) between the respective representative pitches of the question and the response is away from a predetermined reference value (e.g., a consonant interval of a fifth below). Further, a conversation interval detection section (109) is provided for detecting a conversation interval, i.e., a time interval from the end of the question to the start of the response. The evaluation section further evaluates the response to the question on the basis of the detected conversation interval. In this manner, a non-linguistic characteristic of the response to the question (whether an impression given by the response is good or bad) can be evaluated in an objective fashion.