Customer service majorly involves a one-way kind of communication where the organization usually controls the point of interaction through either a call center, helpdesk email address, or even a postal address. The challenges faced by this model are 1) response time (time it takes a customer to get a response about an inquiry they have made) and 2) response rate (rate at which customer inquiries are retrieved and attended to). This paper looks at the use of machine learning algorithms and classifiers, utilizing extractive text summarization techniques for semantic and key phrase extraction of customer queries to facilitate customer response retrieval from a Frequently Asked Questions database. A comparative study of two text summarization approaches (supervised and unsupervised) is carried out by implementing a prototype of an automated agent to respond to customer queries in an electronic media domain. The study illustrates the use of machine learning; text summarization techniques to develop tools that can assist organizations manage their customer interactions effectively and implement robust, efficient, and effective electronic media enabled customer support mechanisms.
