Neural Duplicate Question Detection without Labeled Training Data

TL;DR

This paper introduces two innovative methods for duplicate question detection in community QA forums that do not rely on labeled data, leveraging automatic question generation and weak supervision from question content.

Contribution

The paper presents novel approaches for training duplicate question detection models without labeled data, using automatic question generation and weak supervision from question titles and bodies.

Findings

Both methods outperform traditional supervised approaches in various scenarios.

The approaches effectively utilize large amounts of unlabeled data from cQA forums.

Weak supervision from question content also benefits answer selection models.

Abstract

Supervised training of neural models to duplicate question detection in community Question Answering (cQA) requires large amounts of labeled question pairs, which are costly to obtain. To minimize this cost, recent works thus often used alternative methods, e.g., adversarial domain adaptation. In this work, we propose two novel methods: (1) the automatic generation of duplicate questions, and (2) weak supervision using the title and body of a question. We show that both can achieve improved performances even though they do not require any labeled data. We provide comprehensive comparisons of popular training strategies, which provides important insights on how to best train models in different scenarios. We show that our proposed approaches are more effective in many cases because they can utilize larger amounts of unlabeled data from cQA forums. Finally, we also show that our proposed approach for weak supervision with question title and body information is also an effective method to train cQA answer selection models without direct answer supervision.