Department of Computer Engineering
CS 690 SEMINAR
Incorporation of text mining for customer segmentation and voice record classification
Computer Engineering Department
There has been significant research on collecting and managing customer data in an integrated database for Customer Relationship Management (CRM). Companies now seek to enrich their customer information base by utilizing newly available information channels (e.g. publically available social media information, call center records). In this work, we develop a framework that incorporates text mining to manage a wide variety of market segments for customer segmentation and generate models to classify call center records. Our framework can process categorical and/or text based information associated with the pages liked or POIs checked-in at by the customers in online social networks. It assigns customers into one or more market segments (e.g. sports fans, technology curious, car lovers, frequent shoppers, etc.) using two approaches: (i) dictionary-based relevance scoring and (ii) supervised text-based segmentation. With the help of a supervised text-based prediction, the framework is able to construct the segments even if there is no categorical information available. Our solution incorporates a clustered set of deep learned word vectors to overcome the negative effect of the large number of words present in the learning set. Experimental results confirm that the proposed supervised text-based segmentation approach provides accurate results even when no categorical information is provided. On the other hand, we utilize supervised text-based classification to improve the process of positive/negative call center record detection which is confirmed by experimental results.
DATE: 31 October, 2016, Monday @ 16:00