skip to content

Department of Computer Science and Technology

Date: 
Friday, 6 December, 2019 - 12:00 to 13:00
Speaker: 
Angela Fan (Facebook)
Venue: 
FW26, Computer Laboratory
Abstract: 

We will discuss long-form question answering, a task requiring elaborate and in-depth answers to open-ended questions. The dataset comprises 270K threads from the Reddit forum ``Explain Like I'm Five'' (ELI5) where an online community provides answers to questions which are comprehensible by five year olds. Compared to existing datasets, ELI5 comprises diverse questions requiring multi-sentence answers. We provide a large set of web documents to help answer the question. Automatic and human evaluations show that an abstractive model trained with a multi-task objective outperforms conventional Seq2Seq, language modeling, as well as a strong extractive baseline. However, our best model is still far from human performance since raters prefer gold responses in over 86% of cases, leaving ample opportunity for future improvement. In subsequent work, we propose constructing a local graph structured knowledge base for each query, which compresses the web search information and reduces redundancy. We show that by linearizing the graph into a structured input sequence, models can encode the graph representations within a standard Sequence-to-Sequence setting. We apply this approach to long form question answering. By feeding graph representations as input, we can achieve better performance than using retrieved text portions.

Series: 
NLIP Seminar Series

Upcoming seminars