FAQ-based Question Answering via Word Alignment

TL;DR

This paper introduces a word-alignment-based neural approach for FAQ question answering, utilizing lexical features and a learning-to-rank method, achieving superior performance across multiple languages and tasks.

Contribution

It presents a novel combination of word alignment, bootstrap feature extraction, and learning-to-rank for improved FAQ-based question answering.

Findings

Question similarity model outperforms baselines

Sparse features improve accuracy by 5%

Learning-to-rank surpasses traditional methods

Abstract

In this paper, we propose a novel word-alignment-based method to solve the FAQ-based question answering task. First, we employ a neural network model to calculate question similarity, where the word alignment between two questions is used for extracting features. Second, we design a bootstrap-based feature extraction method to extract a small set of effective lexical features. Third, we propose a learning-to-rank algorithm to train parameters more suitable for the ranking tasks. Experimental results, conducted on three languages (English, Spanish and Japanese), demonstrate that the question similarity model is more effective than baseline systems, the sparse features bring 5% improvements on top-1 accuracy, and the learning-to-rank algorithm works significantly better than the traditional method. We further evaluate our method on the answer sentence selection task. Our method outperforms all the previous systems on the standard TREC data set.