PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval
Sophia Althammer · Zenodo (CERN European Organization for Nuclear Research) · 2021
Here the models described in the publication "PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval" are uploaded. The models are the DPR models denoted as LegalBERT_doc and LegalBERT_para in the paper. The dense retriever models can be loaded with the DPR libary from Facebook Research (https://github.com/facebookresearch/DPR). LegalBERT_para is the dense passage retrieval encoder based on LegalBERT and trained on the paragraph-level labels of COLIEE Task 2 data. LegalBERT_doc is the dense passage retrieval encoder based on LegalBERT and trained on the paragraph-level and document-level labels of COLIEE Task 1&2 data.