Collective Relevance Labeling for Passage Retrieval
This work addresses the need for high-quality labels in IR, offering a more efficient solution for practitioners, though it appears incremental as it builds on existing distillation techniques.
The paper tackles the problem of sparse relevance labels in deep learning for Information Retrieval by proposing a knowledge distillation approach that uses a simple, efficient teacher model to outperform state-of-the-art methods, achieving up to 8x faster training while improving ranking distillation.
Deep learning for Information Retrieval (IR) requires a large amount of high-quality query-document relevance labels, but such labels are inherently sparse. Label smoothing redistributes some observed probability mass over unobserved instances, often uniformly, uninformed of the true distribution. In contrast, we propose knowledge distillation for informed labeling, without incurring high computation overheads at evaluation time. Our contribution is designing a simple but efficient teacher model which utilizes collective knowledge, to outperform state-of-the-arts distilled from a more complex teacher model. Specifically, we train up to x8 faster than the state-of-the-art teacher, while distilling the rankings better. Our code is publicly available at https://github.com/jihyukkim-nlp/CollectiveKD