Multi-Perspective Context Matching for Machine Comprehension

TL;DR

This paper introduces a Multi-Perspective Context Matching model for machine comprehension that directly predicts answer spans in passages, leveraging the SQuAD dataset to improve realism and evaluation of MC techniques.

Contribution

It proposes an end-to-end MC model that matches context from multiple perspectives and adjusts word embeddings based on question relevance, advancing current methods.

Findings

Achieves competitive results on SQuAD leaderboard.

Effectively matches context from multiple perspectives.

Utilizes weighted word embeddings for better relevance modeling.

Abstract

Previous machine comprehension (MC) datasets are either too small to train end-to-end deep learning models, or not difficult enough to evaluate the ability of current MC techniques. The newly released SQuAD dataset alleviates these limitations, and gives us a chance to develop more realistic MC models. Based on this dataset, we propose a Multi-Perspective Context Matching (MPCM) model, which is an end-to-end system that directly predicts the answer beginning and ending points in a passage. Our model first adjusts each word-embedding vector in the passage by multiplying a relevancy weight computed against the question. Then, we encode the question and weighted passage by using bi-directional LSTMs. For each point in the passage, our model matches the context of this point against the encoded question from multiple perspectives and produces a matching vector. Given those matched vectors, we employ another bi-directional LSTM to aggregate all the information and predict the beginning and ending points. Experimental result on the test set of SQuAD shows that our model achieves a competitive result on the leaderboard.