Time: 10:30 am
Location: Via Sommarive 5 - Polo Ferrari 1, Room A203
- Giuseppe "Pino" Di Fabbrizio, VUI
Recent advances in machine learning and natural language processing brought the whole research field on spoken dialog systems back on the main stage. Personal assistants such as Siri, Alexa, and Google Home are now an active part of consumers’ life providing natural and hands-free access to information and services. However, building a conversation agent is still an art and typically requires a level of expertise which is, in most of the cases, out-of-reach for the large majority of small- and medium-sized companies. Such a complex process seriously hinders the adoption of conversational systems by the enterprise digital channels limiting the opportunities for service automation and churn reduction. In this talk, we illustrate machine learning methods that have been successfully applied at web-scale to optimize e-commerce challenges. We also discuss the scalability limitations related to conversational systems and propose techniques to increase data annotation throughput and automatically discover the structure of task-oriented dialogs which are at the very core of statistical modeling and policy learning.
About the speaker
Giuseppe "Pino" Di Fabbrizio is co-founder and CTO of VUI, Inc., the most advanced voice platform for enterprises. Prior to VUI, Pino was a principal research scientist and a group leader in the Rakuten machine learning and natural language processing research team in Boston. Before joining Rakuten, he was a senior research scientist with the Amazon Alexa team in Cambridge, MA. Prior to Amazon, Pino was a lead research scientist at AT&T Labs - Research. His research interests include information extraction, information retrieval, natural language understanding, natural language generation, opinion mining, sentiment analysis, conversational agents, text summarization, multimodal and speech system architectures. He was one of the early team members on the Amazon Echo device project and a main contributor for natural language understanding modeling, new domains expansion, and context dialog modeling. When at AT&T, he lead the productization of the cloud-based AT&T Speech APIs and the development of VoiceTone, a Fortune 500 enterprise customer care automation platform which pioneered speech recognition and natural language processing technology at scale. He published more than 70 papers and was awarded 27 patents. As a senior IEEE member, he regularly contributes to international scientific committees and editorial boards. Di Fabbrizio served as technical chair for Interspeech 2011 and as general co-chair of the IEEE ASRU 2015 workshop.
Contact Person for this Seminar: Prof. giuseppe.riccardi [at] unitn.it (Giuseppe Riccardi)