Explainable Conversational Question Answering over Heterogeneous Sources

Philipp Christmann

Abstract: State-of-the-art conversational question answering (ConvQA) operates over homogeneous sources of information: either a knowledge base (KB), or a text corpus, or a collection of tables. This inherently limits the answer coverage of ConvQA systems. Therefore, during my PhD, we would like to tap into heterogeneous sources for answering conversational questions. Further, we plan to investigate the explainability of such ConvQA systems, to identify what helps users in understanding the answer derivation process.