Abhijit Anand

Abhijit Anand

PhD Student at L3S Research Center, working on docuemnt ranking using laguage models.

Publications

NUMTEMP: A real-world benchmark to verify claims with statistical and temporal expressions

Published in CoRR 24, 2024

Automated fact checking has gained immense interest to tackle the growing misinformation in the digital era. Existing systems primarily focus on synthetic claims on Wikipedia, and noteworthy progress has also been made on real-world claims. In this work, we release Numtemp, a diverse, multi-domain dataset focused exclusively on numerical claims, encompassing temporal, statistical and diverse aspects with fine-grained metadata and an evidence collection without leakage. This addresses the challenge of verifying real-world numerical claims, which are complex and often lack precise information, not addressed by existing works that mainly focus on synthetic claims. We evaluate and quantify the limitations of existing solutions for the task of verifying numerical claims.

Download Paper

The Surprising Effectiveness of Rankers Trained on Expanded Queries

Published in SIGIR 24, 2024

This study addresses the challenge of ranking hard queries in text systems by enriching them with relevant documents and fine-tuning a specialized ranker for these queries. By combining relevance scores from the specialized and base rankers, along with query performance scores, the method achieves significant improvements of up to 25% on passage ranking and up to 48.4% on document ranking tasks compared to baseline approaches, outperforming even state-of-the-art models.

Download Paper

Data Augmentation for Sample Efficient and Robust Document Ranking

Published in ACM Transactions on Information Systems, 2024

The paper proposes data augmentation methods to boost the performance of contextual ranking models, which often demand substantial data for fine-tuning. Through supervised and unsupervised augmentation, relevant document segments in query-document pairs are utilized to enhance sample efficiency, particularly with limited training data. By adapting contrastive losses, the model effectively leverages augmented data, yielding a more robust and effective ranking model for document ranking tasks.

Download Paper

Supervised contrastive learning approach for contextual ranking

Published in ICTIR 22, 2022

This paper proposes a simple yet effective method to improve ranking performance on smaller datasets using supervised contrastive learning for the document ranking problem. We perform data augmentation by creating training data using parts of the relevant documents in the query-document pairs. We then use a supervised contrastive learning objective to learn an effective ranking model from the augmented dataset. Our experiments on subsets of the TREC-DL dataset show that, although data augmentation leads to an increasing the training data sizes, it does not necessarily improve the performance using existing pointwise or pairwise training

Download Paper