← smac.pub home

PARADE: Passage Representation Aggregation for Document Reranking

pdf bibtex code non-refereed

See revised version, published in TOIS 2023 link

Authors: Canjia Li, Andrew Yates, Sean MacAvaney, Ben He, Yingfei Sun

Appeared in: arXiv

DBLP journals/corr/abs-2008-09093 arXiv 2008.09093v2 Semantic Scholar afed54533ecc624cb5e0241172268c6188ded20c smac.pub arxiv2020-parade


We present PARADE, an end-to-end Transformer-based model that considers document-level context for document reranking. PARADE leverages passage-level relevance representations to predict a document relevance score, overcoming the limitations of previous approaches that perform inference on passages independently. Experiments on two ad-hoc retrieval benchmarks demonstrate PARADE's effectiveness over such methods. We conduct extensive analyses on PARADE's efficiency, highlighting several strategies for improving it. When combined with knowledge distillation, a PARADE model with 72\% fewer parameters achieves effectiveness competitive with previous approaches using BERT-Base.

BibTeX @article{li:arxiv2020-parade, author = {Li, Canjia and Yates, Andrew and MacAvaney, Sean and He, Ben and Sun, Yingfei}, title = {PARADE: Passage Representation Aggregation for Document Reranking}, year = {2020}, url = {https://arxiv.org/abs/2008.09093v2}, journal = {arXiv}, volume = {abs/2008.09093} }