Automatic Generation of Domain Models for Call-Centers from Noisy Transcriptions

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 737–744, Sydney, July 2006. c⃝2006 Association for Computational Linguistics Automatic Generation of Domain Models for Call Centers from Noisy Transcriptions Shourya Roy and L Venkata Subramaniam IBM Research India Research Lab IIT Delhi, Block-1 New Delhi 110016 India rshourya,lvsubram@in.ibm.com Abstract Call centers handle customer queries from various domains such as computer sales and support, mo- bile phones, car rental, etc. Each such domain generally has a domain model which is essential to handle customer complaints. These models contain common problem categories, typical cus- tomer issues and their solutions, greeting styles. Currently these models are manually created over time. Towards this, we propose an unsupervised technique to generate domain models automati- cally from call transcriptions. We use a state of the art Automatic Speech Recognition system to transcribe the calls between agents and customers, which still results in high word error rates (40%) and show that even from these noisy transcrip- tions of calls we can automatically build a domain model. The domain model is comprised of pri- marily a topic taxonomy where every node is char- acterized by topic(s), typical Questions-Answers (Q&As), typical actions and call statistics. We show how such a domain model can be used for topic identification of unseen calls. We also pro- pose applications for aiding agents while handling calls and for agent monitoring based on the do- main model. 1 Introduction Call center is a general term for help desks, infor- mation lines and customer service centers. Many companies today operate call centers to handle customer issues. It includes dialog-based (both voice and online chat) and email support a user receives from a professional agent. Call centers have become a central focus of most companies as they allow them to be in direct contact with their customers to solve product-related and services- related issues and also for grievance redress. A typical call center agent handles over a hundred calls in a day. Gigabytes of data is produced ev- ery day in the form of speech audio, speech tran- scripts, email, etc. This data is valuable for doing analysis at many levels, e.g., to obtain statistics about the type of problems and issues associated with different products and services. This data can also be used to evaluate agents and train them to improve their performance. Today’s call centers handle a wide variety of do- mains such as computer sales and support, mobile phones and apparels. To analyze the calls in any domain, analysts need to identify the key issues in the domain. Further, there may be variations within a domain, say mobile phones, based on the service providers. The analysts generate a domain model through inspection of the call records (au- dio, transcripts and emails). Such a model can in- clude a listing of the call categories, types of prob- lems solved in each category, listing of the cus- tomer issues, typical questions-answers, appropri- ate call opening and closing …

Content truncated to 500 words