Text as Data

CHAPTER 8Speech to Text

INTRODUCTION

Researchers have noted that audio files can be preprocessed into an unstructured data stream that can then be used in much the same way as a regular text corpus for analytic processing.ⁱ There is a large market potential and consequently a growing demand for solutions in this area. To illustrate the speech-to-text-to-analytics process, we present a case study that processes consumer audio feedback.

PROCESSING AUDIO FEEDBACK

While it has become easier for consumers to provide feedback to the producer in various forms, one problem for businesses occurs where there is no further information present about the provider of the feedback. Since the provider may not login to a portal or online platform – where fields like name, age, sex are included – businesses may not have access to this information. In this example, we discuss how audio analytics can be used to derive and establish a persona around the feedback provider by predicting sex, age group, ethnicity, and other categorical information. Not only will this help businesses in understanding the demographics of the providers of the feedback but will also help correlate this information to the comments received. The end product will help businesses craft better, more fine-tuned strategies.

Although this approach is relevant in extracting information from textural data in general, in this use case, we are going to explore methods that enable us to extract information from audio data. We are ...