Question Answering over Tables Extracted from Documents

Recent developments in transformer based architectures enabled developing intelligent systems achieving above-human performance on question answering (QA) benchmarks. A natural extension of this research direction is to develop models that can answer questions over tables. In this talk, we will focus on alternative approaches to tackle the problem of finding relevant cell values of a table for a given natural language question.

Mustafa Canim

Mustafa Canim is a Research Scientist at IBM T.J. Watson Research Center. Over the last ten years, he worked with many companies including Airbus, Siemens, KPMG, Repsol (as well as government funded projects) providing AI, NLP and Big Data solutions to the problems enterprise customers are facing. He is currently leading an AI based technical project across the globe, developing a question answering platform for domain specific corpuses. He holds over 40 patents worldwide and published over 30 papers in refereed conferences, including Data Management, Security, Big Data and AI areas.