Data Augmentation for Sample Efficient and Robust Document Ranking

authored by: Abhijit Anand, Jurek Leonhardt, Jaspreet Singh, Koustav Rudra, Avishek Anand
Abstract: Contextual ranking models have delivered impressive performance improvements over classical models in the document ranking task. However, these highly over-parameterized models tend to be data-hungry and require large amounts of data even for fine-tuning. In this article, we propose data-augmentation methods for effective and robust ranking performance. One of the key benefits of using data augmentation is in achieving sample efficiency or learning effectively when we have only a small amount of training data. We propose supervised and unsupervised data augmentation schemes by creating training data using parts of the relevant documents in the query-document pairs. We then adapt a family of contrastive losses for the document ranking task that can exploit the augmented data to learn an effective ranking model. Our extensive experiments on subsets of the MS MARCO and TREC-DL test sets show that data augmentation, along with the ranking-adapted contrastive losses, results in performance improvements under most dataset sizes. Apart from sample efficiency, we conclusively show that data augmentation results in robust models when transferred to out-of-domain benchmarks. Our performance improvements in in-domain and more prominently in out-of-domain benchmarks show that augmentation regularizes the ranking model and improves its robustness and generalization capability.
Organisation(s): L3S Research Centre
External Organisation(s): Delft University of Technology
Indian Institute of Technology Kharagpur (IITKGP)
Type: Article
Journal: ACM Transactions on Information Systems
Volume: 42
No. of pages: 29
ISSN: 1046-8188
Publication date: 29.04.2024
Publication status: Published
Peer reviewed: Yes
ASJC Scopus subject areas: Information Systems, General Business,Management and Accounting, Computer Science Applications
Electronic version(s): https://doi.org/10.48550/arXiv.2311.15426 (Access: Open)
https://doi.org/10.1145/3634911 (Access: Open)

BibTeX

@article{1b737d65f5ee4c00bb185f3743eb81a4,
title = "Data Augmentation for Sample Efficient and Robust Document Ranking",
abstract = "Contextual ranking models have delivered impressive performance improvements over classical models in the document ranking task. However, these highly over-parameterized models tend to be data-hungry and require large amounts of data even for fine-tuning. In this article, we propose data-augmentation methods for effective and robust ranking performance. One of the key benefits of using data augmentation is in achieving sample efficiency or learning effectively when we have only a small amount of training data. We propose supervised and unsupervised data augmentation schemes by creating training data using parts of the relevant documents in the query-document pairs. We then adapt a family of contrastive losses for the document ranking task that can exploit the augmented data to learn an effective ranking model. Our extensive experiments on subsets of the MS MARCO and TREC-DL test sets show that data augmentation, along with the ranking-adapted contrastive losses, results in performance improvements under most dataset sizes. Apart from sample efficiency, we conclusively show that data augmentation results in robust models when transferred to out-of-domain benchmarks. Our performance improvements in in-domain and more prominently in out-of-domain benchmarks show that augmentation regularizes the ranking model and improves its robustness and generalization capability.",
keywords = "contrastive loss, data augmentation, document ranking, Information retrieval, interpolation, IR, ranking, ranking performance",
author = "Abhijit Anand and Jurek Leonhardt and Jaspreet Singh and Koustav Rudra and Avishek Anand",
note = "Publisher Copyright: {\textcopyright} 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.",
year = "2024",
month = apr,
day = "29",
doi = "10.48550/arXiv.2311.15426",
language = "English",
volume = "42",
journal = "ACM Transactions on Information Systems",
issn = "1046-8188",
publisher = "Association for Computing Machinery (ACM)",
number = "5",
}