Exploiting Explicit Paths for Multi-hop Reading Comprehension
We focus on the task of multi-hop reading comprehension where a system is required to reason over a chain of multiple facts, distributed across multiple passages, to answer a question. Inspired by graph-based reasoning, we present a path-based reasoning approach for textual reading comprehension. It operates by generating potential paths across multiple passages, extracting implicit relations along this path, and composing them to encode each path. The proposed model achieves a 2.3 state-of-the-art and, as a side-effect, is also able to explain its reasoning through explicit paths of sentences.
READ FULL TEXT