CNN Agent Behavioral Analytics for Call Center Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating agent performance in call centers are manual, time-consuming, and subjective, making it inefficient to assess agents across an enterprise for workforce optimization and targeted coaching.
Innovation Solution
A system and method using a trained convolutional neural network (CNN) to generate behavioral metrics by transcribing calls and extracting text-based, sentiment-based, and prosody-based features, which are then used to produce behavioral labels for agents, enabling automated evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation by supervisors is used, then agent performance can be assessed, but the process is time-consuming and inefficient
Solution Approach 1:
The patent replaces the mechanical manual evaluation system with an automated machine learning system. Supervisors no longer need to manually listen to and evaluate calls; instead, a trained machine learning model automatically analyzes call transcriptions and generates performance metrics, thereby eliminating time loss while maintaining evaluation accuracy.
Solution Approach 2:
The system enables self-service evaluation where the machine learning model autonomously performs the evaluation task without human intervention. The model processes call data, extracts features, and generates performance assessments automatically, freeing supervisors from this repetitive task while maintaining consistent evaluation standards.
2Measurement precision
If manual evaluation by supervisors is used, then agent performance can be assessed, but the process is highly subjective
Solution Approach 1:
The patent replaces the subjective human judgment mechanism with an objective machine learning-based evaluation system. The model applies consistent criteria across all evaluations, eliminating subjectivity and bias inherent in manual supervisor assessments, while the systematic approach manages the complexity through automated processing.
Solution Approach 2:
The system transforms the evaluation process by changing from qualitative subjective judgments to quantitative objective metrics. The machine learning model extracts measurable parameters from call transcriptions and applies consistent scoring criteria, converting subjective evaluation into objective, data-driven assessments that reduce bias while managing complexity through standardization.
3Productivity
If automated evaluation using machine learning is implemented, then evaluation efficiency is improved, but system complexity increases
Solution Approach 1:
The patent segments the evaluation system into distinct functional modules: call transcription generation, feature extraction, machine learning model processing, and result generation. This modular segmentation manages system complexity by breaking down the automated evaluation process into manageable, independent components that can be developed and maintained separately while achieving high evaluation efficiency.
Solution Approach 2:
The system introduces intermediate processing layers including call transcription as an intermediary between the original call data and the evaluation model, and feature extraction as a mediator that transforms raw transcription data into meaningful inputs for the machine learning model. These intermediaries simplify the overall system complexity by preprocessing data into standardized formats that the evaluation model can efficiently process.
Data Source
AI summary
A system and method for generating an agent behavioral analytics including transcribing an incoming call to produce a call transcription; and using a trained convolutional neural network (CNN) to produce behavioral labels for the agent in the incoming call for behavioral metrics, based on the call transcription. The CNN may include an embedding layer to convert words in the call transcription into vectors in a word embedding space; a convolution layer to perform a plurality of convolutions on the vectors and to generate vectors of features; a pooling layer to concatenate the vectors of features to a single vector by taking a maximum of each feature generated by the convolution layer; and a classification layer to produce grades of the agent in the incoming call for the set of attributes or behavioral metrics, based on the single vector generated by the pooling layer.