Testing Information Extraction Systems with Audible

2017.09.28-NYCML-LoRes-123 (1).jpg

With the goal of improving the Audible customer experience, Audible collaborated with a computer science team from Columbia University to experiment with open-source natural language processing tools. How can we evaluate named entity recognition systems for the text of the audiobooks? How can open-source computational frameworks that can identify people, places, and organizations help Audible better understand the content and context of an audiobook? The team shared their best practices for how data science can influence workflows for tagging and classifying content for better search, discovery and recommendation of audiobooks. 

Participating researchers from Columbia University included: 

Ansaf Salleb-Aouissi
Computer Science Lecturer in Discipline

Travis Riddle
Psychology Department Postdoctoral Research

Jaewan Bahk
Math/Computer Science Undergraduate