Contact Center Sentiment Analysis With Adaptive Lexicons
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sentiment analysis methods struggle with accurately analyzing multi-modal, two-sided, and lengthy interactions in contact centers, which often include informal speech, noise, and sentiment shifts, requiring domain-specific models without sufficient labeled data.
Innovation Solution
A method involving training with generic and domain-specific lexicons, neutral factoring, handling sentiment shifters, pruning, and adaptive segmentation to create a customized sentiment analysis model for contact centers, using machine learning and lexicon integration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If domain-specific models are used for contact center sentiment analysis, then analysis accuracy is improved, but data requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training sentiment analysis models on large amounts of generic domain data before fine-tuning on contact center specific data. This pre-training establishes foundational sentiment understanding that reduces the amount of domain-specific labeled data needed for effective performance.
Solution Approach 2:
The patent merges generic domain sentiment analysis capabilities with contact center specific domain knowledge by combining pre-trained models with domain-specific lexicons and training data. This combination allows the system to leverage both general sentiment understanding and domain-specific nuances with reduced data requirements.
2Loss of information
If multi-modal and lengthy interactions are analyzed, then comprehensive sentiment understanding is improved, but processing complexity increases
Solution Approach 1:
The patent applies segmentation by dividing lengthy contact center interactions into smaller segments or chunks that can be processed individually. This segmentation maintains comprehensive sentiment understanding by analyzing each segment while reducing the overall processing complexity through manageable portions.
Solution Approach 2:
The patent applies dynamics by adapting the analysis approach based on the characteristics of different interaction segments. The system dynamically adjusts processing parameters, segment boundaries, and analysis depth according to the specific properties of each portion of the multi-modal interaction.
3Reliability
If noise and informal speech are handled, then sentiment analysis robustness is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by implementing noise filtering and informal speech normalization as pre-processing steps before main sentiment analysis. This preliminary handling of noisy elements improves robustness while allowing the main analysis to focus on cleaned data, reducing overall processing time.
Solution Approach 2:
The patent extracts and separates noise elements and informal speech patterns from the main content for specialized handling. By taking out these problematic elements for separate processing, the system improves robustness against noise while preventing them from slowing down the main sentiment analysis pipeline.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Methods, systems, and computer program product for automatically performing sentiment analysis on texts, such as telephone call transcripts and electronic written communications. Disclosed techniques include, inter alia, lexicon training, handling of negations and shifters, pruning of lexicons, confidence calculation for token orientation, supervised customization, lexicon mixing, and adaptive segmentation.