MQDD: Pre-training of Multimodal Question Duplicity Detection for Software Engineering Domain

TL;DR

This paper introduces MQDD, a multimodal model trained on Stack Overflow data to improve duplicate question detection in software engineering, utilizing new learning objectives and releasing relevant datasets.

Contribution

The paper presents a novel multimodal pre-training pipeline with new learning objectives for duplicate detection in software engineering questions.

Findings

MQDD effectively detects duplicate questions on Stack Overflow.

The model outperforms existing methods in duplicate detection accuracy.

Two new datasets support further research in this domain.

Abstract

This work proposes a new pipeline for leveraging data collected on the Stack Overflow website for pre-training a multimodal model for searching duplicates on question answering websites. Our multimodal model is trained on question descriptions and source codes in multiple programming languages. We design two new learning objectives to improve duplicate detection capabilities. The result of this work is a mature, fine-tuned Multimodal Question Duplicity Detection (MQDD) model, ready to be integrated into a Stack Overflow search system, where it can help users find answers for already answered questions. Alongside the MQDD model, we release two datasets related to the software engineering domain. The first Stack Overflow Dataset (SOD) represents a massive corpus of paired questions and answers. The second Stack Overflow Duplicity Dataset (SODD) contains data for training duplicate detection models.