Speech Section Classification Using Type-Based Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for classifying speech sections in dialogues face challenges due to the difficulty in accurately labeling short speeches in natural conversations and the inclusion of speeches that do not contribute to classification.
Innovation Solution
A speech section classification device and method that estimate speech sections and types from speech text data, using a classification rule based on the estimated speech types to accurately classify speech sections, even when including non-contributory speeches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is performed using assigned labels for each speech, then a classification model can be generated, but speeches that do not contribute to classification reduce accuracy
Solution Approach 1:
The patent extracts and removes speeches that do not contribute to classification from the input speech data before performing classification. By identifying and eliminating irrelevant speeches (such as filler words, repetitions, or non-informative utterances), the system improves classification accuracy by ensuring that only meaningful speeches are processed by the classification model.
Solution Approach 2:
The patent applies different processing qualities to different speeches within the dialogue. Instead of uniformly processing all speeches, the system identifies speeches with different characteristics and applies appropriate classification or filtering based on their local quality and contribution potential, thereby improving overall classification precision.
2Productivity
If all speeches are applied to the classifier, then complete speech sections can be processed, but classification accuracy decreases due to non-contributory speeches
Solution Approach 1:
The system extracts and removes non-contributory speeches from the speech section before classification, maintaining processing completeness while improving accuracy by eliminating irrelevant data points that would otherwise dilute the classification results.
Solution Approach 2:
The patent performs preliminary processing of speeches before classification by identifying and removing non-contributory speeches in advance. This preliminary action ensures that the classification model receives only relevant input, improving accuracy without compromising the completeness of meaningful speech sections.
3Measurement precision
If labels are assigned to each short speech in natural conversation, then individual speech classification is possible, but labeling difficulty increases due to speech brevity
Solution Approach 1:
The system employs automatic label assignment through machine learning models that self-service the labeling process. Instead of requiring manual labeling of each short speech, the model automatically assigns labels based on learned patterns from training data, making the process scalable and reducing the difficulty of labeling brief utterances.
Solution Approach 2:
The patent changes the approach from manual labeling to automated machine learning-based labeling, fundamentally altering the parameter of label assignment from human effort to algorithmic processing. This parameter change enables efficient handling of short speeches by leveraging patterns learned from larger datasets.
Data Source
AI summary
A speech section classification device includes: a speech section estimation unit that estimates a speech section from speech text data including speeches of two or more people; a speech type estimation unit that estimates a speech type of each speech included in the speech section estimated by the speech section estimation unit; and a speech section classification unit that classifies the speech section estimated by the speech section estimation unit, using the speech type of each speech estimated by the speech type estimation unit and a speech section classification rule determined in advance as a rule for classifying speech sections on the basis of the speech type.