Automatic Interaction Quality Evaluation Using Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quality monitoring systems in contact centers rely heavily on manual evaluations, which are time-consuming and inefficient, leading to only a small fraction of interactions being evaluated, resulting in inadequate feedback and performance trends analysis for agents.
Innovation Solution
A system and method for automatically evaluating interactions using a combination of automatic and manual questions, where automatic features are extracted from interactions to compute an overall evaluation score, including the use of deep neural networks trained on evaluation forms to provide real-time feedback and customized training sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation is used to assess agent performance, then evaluation accuracy is improved, but evaluation efficiency deteriorates
Solution Approach 1:
The system enables automatic self-evaluation of agent interactions through machine learning models that autonomously analyze call recordings, transcribe speech, and generate performance scores without human intervention. This resolves the contradiction by making the evaluation system serve itself, achieving both high efficiency through automation and maintained accuracy through trained ML algorithms.
Solution Approach 2:
The patent replaces the mechanical manual evaluation process with an automated machine learning-based system. The ML model substitutes human evaluators, performing speech recognition, sentiment analysis, and performance scoring automatically. This substitution achieves high efficiency while maintaining accuracy through the sophisticated analytical capabilities of the ML system.
2Reliability
If manual evaluation is used to ensure quality, then evaluation thoroughness is improved, but time consumption deteriorates
Solution Approach 1:
The system enables continuous automatic evaluation of all agent interactions in real-time or near-real-time, eliminating the intermittent nature of manual evaluation. The ML model continuously processes call recordings as they occur, providing ongoing quality assurance without the time delays inherent in manual review processes, thus achieving both thoroughness and speed.
Solution Approach 2:
The evaluation system autonomously performs comprehensive analysis of interactions including speech transcription, sentiment detection, compliance checking, and performance scoring without requiring human time investment. This self-service capability ensures thorough evaluation of all quality aspects while consuming minimal time.
3Productivity
If only a small fraction of interactions are evaluated, then resource efficiency is improved, but feedback completeness deteriorates
Solution Approach 1:
The ML-based evaluation system is designed to universally process and evaluate all types of agent-customer interactions across multiple channels (voice, chat, email). It simultaneously performs multiple evaluation functions including quality assessment, compliance monitoring, sentiment analysis, and performance tracking, enabling comprehensive feedback on 100% of interactions without additional resource burden.
Solution Approach 2:
The system autonomously evaluates every interaction without requiring proportional human resources, using automated ML algorithms to process and analyze complete interaction datasets. This self-service capability ensures no information loss while maintaining resource efficiency through automation.
4Productivity
If automatic evaluation is implemented, then evaluation efficiency is improved, but evaluation complexity deteriorates
Solution Approach 1:
The evaluation system is segmented into distinct functional modules: speech recognition module, sentiment analysis module, compliance checking module, and scoring module. Each module handles a specific aspect of evaluation independently, then integrates results. This segmentation manages complexity by breaking down the automated evaluation process into manageable, specialized components while maintaining high efficiency.
Data Source
AI summary
A method for automatically calculating an overall evaluation score of an interaction includes: receiving, by a processor, an evaluation form, the evaluation form comprising a plurality of automatic questions and a plurality of manual questions; automatically extracting, by a processor, a set of features from the interaction, the set of features comprising answers to the automatic questions without manually generated answers to the manual questions; and computing an overall evaluation score based on the set of features.