Call Transcript Intent Filtering for Accurate Service Summaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generically trained speech-to-text and summarization models lack accuracy for processing specialized customer service calls, limiting their effectiveness in generating reliable call summaries and integrating with other data mining activities.
Innovation Solution
A trained intent model processes call transcripts to determine intent labels for utterances, which are then filtered and processed through a trained summarization model to generate accurate natural language summaries, incorporating speaker identifiers and topic classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If generically trained speech-to-text and summarization models are used, then device complexity is reduced, but manufacturing precision (accuracy) deteriorates
Solution Approach 1:
The system segments the call processing into distinct stages: speech-to-text conversion, utterance extraction, intent classification, filtering, and summarization. Each stage uses specialized models trained for specific purposes, with the intent model filtering utterances before the summarization model processes them, thereby improving overall accuracy without requiring a single overly complex model
Solution Approach 2:
The intent classification model performs preliminary action by classifying and filtering utterances before they are passed to the summarization model. This pre-processing step removes irrelevant or low-value utterances, allowing the summarization model to focus on high-quality input data, thereby improving accuracy while maintaining reasonable complexity
2Productivity
If all utterances are processed through the summarization model, then productivity is improved, but loss of information increases due to including irrelevant data
Solution Approach 1:
The system extracts only the most relevant utterances from the call transcript by using the intent classification model to identify and filter utterances that contain valuable information. This extraction process removes irrelevant or redundant utterances before summarization, preserving information quality while improving processing efficiency by reducing the volume of data requiring full summarization
3Manufacturing precision
If specialized models are trained for accurate call processing, then manufacturing precision improves, but device complexity increases
Solution Approach 1:
The system employs multiple specialized models that each perform a specific function (speech-to-text, intent classification, summarization), but these models work together in a unified pipeline that handles the complete call processing workflow. This multi-functional approach achieves high accuracy through specialization while maintaining manageable complexity through modular architecture and clear division of labor
Data Source
AI summary
Systems and methods for automated summarization of customer service calls in a specialized field, such as medical services and insurance, are disclosed. Call transcripts are processed to determine the utterances and corresponding speakers in a customer service call. A feature vector is generated for each utterance and processed through a trained intent model to assign intent labels. The utterances are then filtered based on the intent labels to generate an extractive summary. The extractive summary is processed through a trained summarization model to generate a natural language summary for the customer service call.