Semi-structured tables are commonly seen in everyday life as one of the most popular and convenient ways to store and organize data. Recently, they have been recognized as a rich knowledge source for the Question Answering (QA) tasks. Unlike relational databases tables, it is more challenging for machines to automatically understand semi-structured tables and use them in downstream tasks. Due to the complexity of Table QA, existing works tried to crack the problem as sub-tasks, i.e., table retrieval and QA over tables. The traditional two-step pipeline has its limitation on performance, mainly due to error propagation. In this talk, I introduce a series of models aiming to fill in the vacancy of end-to-end Table QA. Given any natural language question, the proposed models can efficiently search through a massive table corpus, retrieve the relevant tables, and locate the correct table cells to answer the questions. The models leverage the transformer-based framework with the support of semantic-driven approaches and achieve state-of-the-art performance on recent benchmarks.
Feifei Pan is a Ph.D. candidate in Computer Science at Rensselaer Polytechnic Institute, advised by Prof. Jim Hendler and Prof. Peter Fox. Her recent research focuses on Natural Language Processing (NLP) and Machine Learning (ML), especially table question answering and ML for geoscience data. She has published her work in top-tier conferences and journals, including ACL, NAACL, AAAI, and Nature Communications. She is a recipient of the IBM AI Horizon Fellowship.