PassageReranking Baseline

implentment by pytorch

dataset format

train dataset

• collection.train.sampled.tsv

  pid passage

• queries.train.sampled.tsv

  qid query

• qidpidtriples.train.sampled.tsv

  qid positive_pid negative_pid

validation and test dataset

• msmarco-passagetest2019-43-top1000.tsv

  qid pid query passage

• msmarco-passagetest2020-54-top1000.tsv

  qid pid query passage

train

train.sh

test

evaluate.sh

eval by ndcg

eval.py

or you can eval usr offical offered C++ script

download from https://trec.nist.gov/trec_eval/

tar xzvf trec_eval-9.0.7.tar.gz
cd trec_eval-9.0.7
make
./trec_eval-9.0.7/trec_eval -m qrels model_result

trec_eval.sh