Answer Selection in Community Question Answering Exploiting Knowledge Graph and Context Information

Tracking #: 2675-3889

This paper is currently under review
Golshan Afzali
Heshaam Faili1

Responsible editor: 
Guest Editors DeepL4KGs 2021

Submission type: 
Full Paper
With the increasing popularity of knowledge graphs (KGs), many applications such as sentiment analysis, trend prediction, and question answering use KG for matching the entities mentioned in the text to entities in the KG. Despite the usefulness of commonsense knowledge or factual background knowledge in the KGs, to the best of our knowledge, these KGs have been rarely used for answer selection in community question answering (CQA). In this paper, we propose a novel answer selection method in CQA by using the knowledge embedded in KG. Our method is a deep neural network based model that besides using KG, uses a latent-variable model for learning the representations of the question and answer, by jointly optimizing generative and discriminative objectives. Specifically, the proposed model leverages external background knowledge from KG to help identify entity mentions and their relations. It also uses the question category for producing a context-aware representation for each of the question and answer. Moreover, the model uses variational autoencoders (VAE) in a multi-task learning process with a classifier to produce a class-specific representation for each answer. The experimental results on three widely used datasets demonstrate that the proposed method significantly outperforms all existing models in this field.
Full PDF Version: 
Under Review