Conversational Sentiment Classification With Emotion Strength Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sentiment analysis systems fail to accurately distinguish between neutral and other sentiments in conversational data, often misclassifying emotions and not considering emotion in sentiment analysis, leading to unsatisfactory results and increased manual effort in handling data, especially in call centers and online chat platforms.
Innovation Solution
A sentiment classification system that incorporates emotion analysis, using a deep learning architecture with a sentiment classifier and two emotion classifiers to predict sentiment and emotion strength at the utterance level, providing a dynamic representation of sentiment and emotion flow throughout a conversation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional off-the-shelf sentiment analysis is used, then the system is simple to implement, but it fails to perform accurately enough in the context of conversations
Solution Approach 1:
The patent segments the sentiment analysis task into multiple specialized components: a sentiment classifier for determining overall sentiment polarity, an emotion classifier for detecting specific emotions, and a sentiment strength predictor for measuring intensity. This segmentation allows each component to specialize in one aspect, improving overall accuracy while maintaining manageable complexity through modular design.
Solution Approach 2:
The patent merges multiple classification functions (sentiment classification, emotion detection, and strength prediction) into a unified system that processes conversational data together. By combining these functions that previously operated independently, the system captures the interplay between sentiment, emotion, and intensity, achieving more accurate sentiment analysis than separate systems could provide.
2Measurement precision
If emotion analysis is incorporated into sentiment analysis, then sentiment classification accuracy improves, but system complexity increases
Solution Approach 1:
The patent divides the analysis into distinct modular components: a sentiment classifier that determines polarity, an emotion classifier that detects specific emotions, and a strength predictor that measures intensity. Each module handles a specific aspect independently, allowing them to be developed, trained, and maintained separately while working together to improve overall accuracy.
Solution Approach 2:
The patent introduces an emotion classifier as an intermediary component between raw conversational input and final sentiment classification. This intermediary extracts emotion information that then informs the sentiment analysis, allowing the system to capture emotional nuances that would otherwise be missed, thereby improving accuracy without directly complicating the core sentiment classification logic.
3Productivity
If unweighted sentiment analysis is used, then processing is simple and fast, but it does not rank text segments with negative and positive sentiments, resulting in heavy manual effort
Solution Approach 1:
The patent introduces a sentiment strength parameter that quantifies the intensity of sentiment in addition to the binary positive/negative classification. By changing from a simple categorical output to a continuous strength metric, the system automatically prioritizes high-strength sentiments, enabling efficient identification of critical issues without manual review of all sentiment-labeled texts.
4Loss of information
If sentiment is determined based on overall text in a session, then analysis is simpler, but it fails to identify which specific utterances contribute to negative or positive sentiment evaluation
Solution Approach 1:
The patent applies segmentation at the utterance level, processing and classifying each individual utterance separately rather than analyzing the entire conversation as a single unit. This allows the system to identify which specific utterances drive the overall sentiment, providing actionable insights into problem areas while maintaining manageable complexity through incremental processing of discrete segments.
Data Source
AI summary
A system includes an input interface receiving conversation input from at least one system participant. A natural language processor operates on the conversational input. A sentiment processor operates on the pre-processed input using a sentiment classification model to predict sentiment, the sentiment processor predicting positive, neutral, and negative sentiment for the pre-processed input. An emotion processor receiving the pre-processed input from the natural language processor and a corresponding sentiment from the sentiment processor and operates to predict an emotion class and a strength of the emotion based on the corresponding sentiment. An output mechanism generates sentiment strength for the conversation input based on the emotion class and strength of the emotion.