An emerging recipe for achieving state-of-the-art effectiveness in neuraldocument re-ranking involves utilizing large pre-trained language models -e.g., BERT – to evaluate all individual passages in the document and thenaggregating the outputs by pooling or additional Transformer layers . The proposed Intra-DocumentCascaded Ranking Model (IDCM) leads to over 400% lower query latency by providing essentially the same effectiveness as BERT-based ranking models . This method prunes passages of a candidate document using a less expensive model, called ESM, beforerunning a scoring model that is more expensive and effective, called ETM . Thispruning allows us to only run the ETM model on a smaller set of passages whosesize does not vary by document length. Thisprune allows us to onlyrun the ETm model on smaller sets of passages which does not depend on the size of the document. The proposed IDCM leads to less expensive
Author(s) : Sebastian Hofstätter, Bhaskar Mitra, Hamed Zamani, Nick Craswell, Allan HanburyLinks : PDF - Abstract
Code :
Keywords : document - model - passages - ranking - expensive -