Customer Care Innovation

Quality Assurance in Real Time (QART)

An end-to-end, real-time quality assurance system for contact centers.

Contact centers provide dialogue (both voice and online chat) and email-based support to solve product and services-related issues, queries, and requests. Today, contact centers exploit cost arbitrage through outsourcing and (partial) automation of service delivery processes such as agent assistance tools. At the same time, providing the highest quality of service leading to satisfied customers is also a key objective. The two most commonly employed practices towards this goal, namely Quality Assurance (QA) and Customer Satisfaction analysis (C-Sat), as practiced today, have several shortcomings. Firstly, they reveal outcomes of dialogues and associated reasons always in hindsight, because manual live dialogue monitoring is often impractical. Secondly, it is not uncommon for QA and C-Sat to have conflicting outcomes. For example, an agent and her supervisor may feel that in a dialogue, everything was done perfectly, but the customer’s feedback could still be negative. Moreover, only a small fraction of the contact center workforce is responsible for these processes, so, only a small sample of interactions ends up getting analyzed.

Our Approach

We have developed QART, an automatic, real-time quality assurance system for contact centers. QART performs holistic, multi-faceted analysis on each and every utterance made by customers and agents by bringing together and automating various aspects of manual QA and C-Sat processes. Our researchers have exploited novel features and techniques in natural language processing, built on state-of-the-art machine learning technology, to extract the most relevant information from utterances, which capture behavioral aspects of customer satisfaction. QART's key modules include an incremental dialogue summarizer and an interactive real-time dialogue dashboard, which enables supervisors to obtain glanceable views and on-demand details about potentially problematic dialogues in real-time, enabling seamless interventions. A large number of interactions can be processed simultaneously, thanks to automation, which spurs contact centers to go beyond sampling-based QA.

Business Engagements

Pilot Interest from external BPO customers.

Technical Accomplishments

"Fine-grained Emotion Detection in Contact CenterChat Utterances using a Neural Network Approach," Patent Filed.
"Building Emotion Lexicon with Minimal Supervision using Contextual Information," Patent Filed.
"Multi-task learning of word embedding representations for emotion classification," Patent Filed.
"A method for Handling Figurative Speech for emotion classification in micro-blog texts," Patent Filed.

Scientific Impact

"Fine-grained Emotion Detection in Contact Center Chat Utterances," accepted for publication in the Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), 2017.
"Multi-task Representation Learning for Enhanced Emotion Categorization in Short Text," accepted for publication in the Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), 2017.
"QART: A System for Real-time Holistic Quality Assurance for Contact Center Dialogues," published in the 30th AAAI Conference on Artificial Intelligence (AAAI), 2016.
"QART: A Tool for Quality Assurance in Real-Time in Contact Centers," published in the 25th International Conference on Information and Knowledge Management (CIKM), 2016.