MultiCQA: Zero-Shot Transfer of Self-Supervised Text Matching Models on a Massive Scale

TL;DR

This paper demonstrates that self-supervised text matching models trained on numerous community question answering domains exhibit strong zero-shot transfer capabilities, outperforming traditional baselines and previous state-of-the-art models across multiple benchmarks.

Contribution

It introduces a large-scale self-supervised training approach across 140 domains and explores effective multi-source domain combination methods for zero-shot transfer.

Findings

Models transfer well across diverse domains.

Broad source domain selection improves transfer performance.

Multi-task learning enhances zero-shot transfer results.

Abstract

We study the zero-shot transfer capabilities of text matching models on a massive scale, by self-supervised training on 140 source domains from community question answering forums in English. We investigate the model performances on nine benchmarks of answer selection and question similarity tasks, and show that all 140 models transfer surprisingly well, where the large majority of models substantially outperforms common IR baselines. We also demonstrate that considering a broad selection of source domains is crucial for obtaining the best zero-shot transfer performances, which contrasts the standard procedure that merely relies on the largest and most similar domains. In addition, we extensively study how to best combine multiple source domains. We propose to incorporate self-supervised with supervised multi-task learning on all available source domains. Our best zero-shot transfer model considerably outperforms in-domain BERT and the previous state of the art on six benchmarks. Fine-tuning of our model with in-domain data results in additional large gains and achieves the new state of the art on all nine benchmarks.