Sentiment Analysis of Call Centre Audio Conversations using Text Classification

The field of text mining has evolved over the past few years to help analyze the vast amount of textual resources available online. Text Mining, however, can be used also in various other applications. In this research, we are particularly interested in performing text mining techniques over transcribed audio recordings in order to detect the speakers' emotions. Our work is originally motivated by use cases arising from call centers, but can also have applications in other areas. We describe our overall methodology and present our experimental results for speech-to-text transcription, text classification and text clustering. We also focus on analyzing the effects of using different features selection methods.