Review-based Question Generation with Adaptive Instance Transfer and Augmentation

TL;DR

This paper introduces a novel framework for generating questions from online reviews by leveraging adaptive transfer and augmentation techniques, addressing data scarcity and capturing review aspects to improve question generation quality.

Contribution

It presents an iterative learning framework that uses transfer and augmentation of review-question pairs, incorporating unsupervised feature extraction to enhance question generation from reviews.

Findings

Effective question generation demonstrated across 10 e-commerce categories

Framework improves over baseline models in quality and relevance

Augmentation and transfer strategies significantly enhance training data utility

Abstract

Online reviews provide rich information about products and service, while it remains inefficient for potential consumers to exploit the reviews for fulfilling their specific information need. We propose to explore question generation as a new way of exploiting review information. One major challenge of this task is the lack of review-question pairs for training a neural generation model. We propose an iterative learning framework for handling this challenge via adaptive transfer and augmentation of the training instances with the help of the available user-posed question-answer data. To capture the aspect characteristics in reviews, the augmentation and generation procedures incorporate related features extracted via unsupervised learning. Experiments on data from 10 categories of a popular E-commerce site demonstrate the effectiveness of the framework, as well as the usefulness of the new task.