Enhancing the Ranking Context of Dense Retrieval through Reciprocal Nearest Neighbors
Submission Type: Regular Long Paper
Submission Track: Information Retrieval and Text Mining
Keywords: dense retrieval, reciprocal nearest neighbors, ranking context, contrastive learning, list-wise loss, false negatives, label smoothing, transformers, Large Language Models, information retrieval, natural language processing, deep learning
TL;DR: To address sparse document annotation and false negatives when training dense retrieval models, we introduce evidence-based label smoothing that leverages reciprocal nearest neighbors to improve the ranking context.
Abstract: Sparse annotation poses persistent challenges to training dense retrieval models; for example, it distorts the training signal when unlabeled relevant documents are used spuriously as negatives in contrastive learning. To alleviate this problem, we introduce evidence-based label smoothing, a novel, computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives. To compute the target relevance distribution over candidate documents within the ranking context of a given query, we assign a non-zero relevance probability to those candidates most similar to the ground truth based on the degree of their similarity to the ground-truth document(s).
To estimate relevance we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in post-processing. Through extensive experiments on two large-scale ad hoc text retrieval datasets, we demonstrate that reciprocal nearest neighbors can improve the ranking effectiveness of dense retrieval models, both when used for label smoothing, as well as for reranking. This indicates that by considering relationships between documents and queries beyond simple geometric distance we can effectively enhance the ranking context.
Submission Number: 5303
Loading