Intent-bearing Utterance Identification in Call Center Conversations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Call centers face challenges in efficiently processing large volumes of customer conversations to identify the intent behind calls, which is crucial for customer relationship management, resource allocation, and customer satisfaction, as existing methods are either too broad or fail to accurately pinpoint intent-bearing utterances within conversations.
Innovation Solution
A method and apparatus that determine features for each utterance in a conversation, classify them using a classifier, and select intent-bearing utterances based on these classifications and assigned scores, potentially using conditional random fields and state sequences to maximize cumulative scores for accurate intent identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional broad classification methods are used to process customer conversations, then processing coverage is improved, but measurement precision of intent identification deteriorates
Solution Approach 1:
The patent segments the conversation processing task into multiple levels: first dividing conversations into utterances, then classifying each utterance individually as intent-bearing or non-intent-bearing, and finally aggregating these classifications to determine overall conversation intent. This segmentation allows the system to maintain high measurement precision at the utterance level while achieving comprehensive processing coverage across entire conversations.
Solution Approach 2:
The patent applies different processing qualities to different parts of the conversation data. Specifically, it applies detailed feature extraction and classification to individual utterances (local level) while maintaining an overview of the entire conversation (global level). This local quality approach enables precise intent identification in critical utterances without sacrificing overall processing coverage.
2Measurement precision
If detailed feature analysis is applied to each utterance, then intent identification accuracy is improved, but device complexity increases
Solution Approach 1:
The system segments the complex task of intent identification into manageable components: feature extraction, classification, and aggregation. By processing utterances independently through this segmented pipeline, the system achieves high accuracy without requiring a monolithic complex system, thereby reducing overall device complexity.
Solution Approach 2:
The patent employs a universal classifier that can handle multiple types of features and classify different utterances using the same underlying mechanism. This multi-functionality reduces device complexity by avoiding the need for separate specialized systems for each classification task, while still maintaining high intent identification accuracy through consistent application of the classification logic.
3Measurement precision
If manual intent identification is performed, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The system implements self-service through automated feature extraction and classification of utterances. The classifier automatically determines which utterances are intent-bearing without human intervention, and the aggregation process automatically synthesizes conversation-level intent from utterance-level classifications. This automation maintains measurement precision comparable to manual methods while dramatically improving productivity by processing calls at machine speed.
Solution Approach 2:
The patent replaces the mechanical process of manual intent identification with an automated computational system. Instead of human analysts manually reviewing and classifying utterances, the system uses algorithmic feature extraction and machine learning-based classification to automatically identify intent-bearing utterances, thereby maintaining accuracy while eliminating the productivity bottleneck of manual processing.
4Productivity
If automated classification is implemented across all utterances, then productivity is improved, but measurement precision deteriorates
Solution Approach 1:
The system processes each utterance as a separate segment with its own feature extraction and classification, rather than applying a single bulk classification to the entire conversation. This segmentation allows the automated system to maintain high measurement precision at the utterance level while achieving high productivity through automated processing of all segments in parallel or sequence.
Solution Approach 2:
The patent dynamically adjusts classification parameters and thresholds based on the specific characteristics of each utterance and the overall conversation context. This parameter adaptation enables the automated classification system to maintain high measurement precision across diverse utterance types while preserving the productivity benefits of automation, as the system can optimize its decision criteria for each classification task.
Data Source
AI summary
Identification of an intent of a conversation can be useful for real-time or post-processing purposes. According to example embodiments, a method, and corresponding apparatus of identifying at least one intent-bearing utterance in a conversation, comprises determining at least one feature for each utterance among a subset of utterances of the conversation; classifying each utterance among the subset of utterances, using a classifier, as an intent classification or a non-intent classification based at least in part on a subset of the at least one determined feature; and selecting at least one utterance, with intent classification, as an intent-bearing utterance based at least in part on classification results by the classifier. Through identification of an intent bearing utterance, a call center for example, can provide improved service for callers through, for example, more effective directing of a call to a live agent.