Ultra-High Dimensional Sparse Representations with Binarization for Efficient Text Retrieval

Abstract

The semantic matching capabilities of neural information retrieval can ameliorate synonymy and polysemy problems of symbolic approaches. However, neural models' dense representations are more suitable for re-ranking, due to their inefficiency. Sparse representations, either in symbolic or latent form, are more efficient with an inverted index. Taking the merits of the sparse and dense representations, we propose an ultra-high dimensional (UHD) representation scheme equipped with directly controllable sparsity. UHD's large capacity and minimal noise and interference among the dimensions allow for binarized representations, which are highly efficient for storage and search. Also proposed is a bucketing method, where the embeddings from multiple layers of BERT are selected/merged to represent diverse linguistic aspects. We test our models with MS MARCO and TREC CAR, showing that our models outperforms other sparse models

arXiv commentTo appear at EMNLP 2021

References in corpus (11)

• A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks
• Passage Re-ranking with BERT
• Query expansion techniques for information retrieval: A survey
• Document Expansion by Query Prediction
• Properties of Sparse Distributed Representations and their Application to Hierarchical Temporal Memory
• RocketQA: An Optimized Training Approach to Dense Passage Retrieval for Open-Domain Question Answering
• How Can We Be So Dense? The Benefits of Using Highly Sparse Representations
• RepBERT: Contextualized Text Embeddings for First-Stage Retrieval
• SparTerm: Learning Term-based Sparse Representation for Fast Text Retrieval
• Minimizing FLOPs to Learn Efficient Sparse Representations
• SOLAR: Sparse Orthogonal Learned and Random Embeddings