Group Sparse CNNs for Question Classification with Answer Sets

TL;DR

This paper introduces group sparse CNNs that incorporate answer set information into question classification, significantly improving performance by refining question representations with novel autoencoders.

Contribution

It presents a new group sparse autoencoder and integrates it into CNNs to enhance question classification using answer set data, a novel approach in the field.

Findings

Model outperforms strong baselines on four datasets.

Group sparse autoencoders effectively utilize answer information.

Improved question representations lead to better classification accuracy.

Abstract

Question classification is an important task with wide applications. However, traditional techniques treat questions as general sentences, ignoring the corresponding answer data. In order to consider answer information into question modeling, we first introduce novel group sparse autoencoders which refine question representation by utilizing group information in the answer set. We then propose novel group sparse CNNs which naturally learn question representation with respect to their answers by implanting group sparse autoencoders into traditional CNNs. The proposed model significantly outperform strong baselines on four datasets.