Auto Predictive Customer Feedback from Textual Analysis of Online Chat Logs

Abstract

Text analytics is the process to convert unstructured data into structured or meaningful data for analysis by structuring the input text for deducing patterns and trends to evaluate and interpret the output data. It measures customer opinions, product reviews and feedback for providing sentiment analysis to support a decision making which is based upon facts. Data mining and text analytics when combined with statistics creates predictive intelligence to uncover patterns and relationships for both unstructured and structured data. This research work presents a concept to collect online chat logs for the customer at all timestamps for a chat conversation to auto predict the customer feedback using textual sentiment analysis. Using term document frequency, a matrix is obtained with the word count as an input for performing sentiment analysis on the available set of words. The obtained results deduce positive and negative sentiments for the provided words in a matrix which derives association rules using “Apriori algorithm” among the positive and negative sentiments for the most common set of pre-defined words in a dictionary. These association rules will infer relationships among these sentiments for auto predicting customer feedback via online chat logs, eliminating manual filling of feedback forms.