SimANS: Simple Ambiguous Negatives Sampling for Dense Text Retrieval

10/21/2022
by   Kun Zhou, et al.
0

Sampling proper negatives from a large document pool is vital to effectively train a dense retrieval model. However, existing negative sampling strategies suffer from the uninformative or false negative problem. In this work, we empirically show that according to the measured relevance scores, the negatives ranked around the positives are generally more informative and less likely to be false negatives. Intuitively, these negatives are not too hard (may be false negatives) or too easy (uninformative). They are the ambiguous negatives and need more attention during training. Thus, we propose a simple ambiguous negatives sampling method, SimANS, which incorporates a new sampling probability distribution to sample more ambiguous negatives. Extensive experiments on four public and one industry datasets show the effectiveness of our approach. We made the code and models publicly available in <https://github.com/microsoft/SimXNS>.

READ FULL TEXT
research
08/08/2023

Your Negative May not Be True Negative: Boosting Image-Text Matching with False Negative Elimination

Most existing image-text matching methods adopt triplet loss as the opti...
research
11/08/2022

On Negative Sampling for Contrastive Audio-Text Retrieval

This paper investigates negative sampling for contrastive learning in th...
research
09/12/2022

Hard Negatives or False Negatives: Correcting Pooling Bias in Training Neural Ranking Models

Neural ranking models (NRMs) have become one of the most important techn...
research
05/25/2022

Revisiting DocRED – Addressing the Overlooked False Negative Problem in Relation Extraction

The DocRED dataset is one of the most popular and widely used benchmarks...
research
08/08/2022

UFNRec: Utilizing False Negative Samples for Sequential Recommendation

Sequential recommendation models are primarily optimized to distinguish ...
research
07/25/2020

Bollyrics: Automatic Lyrics Generator for Romanised Hindi

Song lyrics convey a meaningful story in a creative manner with complex ...
research
06/03/2019

Gendered Ambiguous Pronouns Shared Task: Boosting Model Confidence by Evidence Pooling

This paper presents a strong set of results for resolving gendered ambig...

Please sign up or login with your details

Forgot password? Click here to reset