Skip to content

Latest commit

 

History

History
75 lines (55 loc) · 4 KB

regressions-beir-v1.0.0-quora-flat-wp.md

File metadata and controls

75 lines (55 loc) · 4 KB
Raw

Anserini Regressions: BEIR (v1.0.0) — Quora

This page documents BM25 regression experiments for BEIR (v1.0.0) — Quora. These experiments index the corpus in a "flat" manner, by concatenating the "title" and "text" into the "contents" field. All the documents and queries are pre-tokenized with bert-base-uncased tokenizer.

The exact configurations for these regressions are stored in this YAML file. Note that this page is automatically generated from this template as part of Anserini's regression pipeline, so do not modify this page directly; modify the template instead.

From one of our Waterloo servers (e.g., orca), the following command will perform the complete regression, end to end:

python src/main/python/run_regression.py --index --verify --search --regression beir-v1.0.0-quora-flat-wp

All the BEIR corpora, pre-tokenized with the bert-base-uncased tokenizer, are available for download:

wget https://rgw.cs.uwaterloo.ca/pyserini/data/beir-v1.0.0-corpus-wp.tar -P collections/
tar xvf collections/beir-v1.0.0-corpus-wp.tar -C collections/

The tarball is 13 GB and has MD5 checksum 3cf8f3dcdcadd49362965dd4466e6ff2. After download and unpacking the corpora, the run_regression.py command above should work without any issue.

Indexing

Typical indexing command:

bin/run.sh io.anserini.index.IndexCollection \
  -collection BeirFlatCollection \
  -input /path/to/beir-v1.0.0-quora-flat-wp \
  -generator DefaultLuceneDocumentGenerator \
  -index indexes/lucene-index.beir-v1.0.0-quora-flat-wp/ \
  -threads 1 -storePositions -storeDocvectors -storeRaw -pretokenized \
  >& logs/log.beir-v1.0.0-quora-flat-wp &

For additional details, see explanation of common indexing options.

Retrieval

Topics and qrels are stored here, which is linked to the Anserini repo as a submodule.

After indexing has completed, you should be able to perform retrieval as follows:

bin/run.sh io.anserini.search.SearchCollection \
  -index indexes/lucene-index.beir-v1.0.0-quora-flat-wp/ \
  -topics tools/topics-and-qrels/topics.beir-v1.0.0-quora.test.wp.tsv.gz \
  -topicReader TsvString \
  -output runs/run.beir-v1.0.0-quora-flat-wp.bm25.topics.beir-v1.0.0-quora.test.wp.txt \
  -bm25 -removeQuery -pretokenized &

Evaluation can be performed using trec_eval:

bin/trec_eval -c -m ndcg_cut.10 tools/topics-and-qrels/qrels.beir-v1.0.0-quora.test.txt runs/run.beir-v1.0.0-quora-flat-wp.bm25.topics.beir-v1.0.0-quora.test.wp.txt
bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.beir-v1.0.0-quora.test.txt runs/run.beir-v1.0.0-quora-flat-wp.bm25.topics.beir-v1.0.0-quora.test.wp.txt
bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.beir-v1.0.0-quora.test.txt runs/run.beir-v1.0.0-quora-flat-wp.bm25.topics.beir-v1.0.0-quora.test.wp.txt

Effectiveness

With the above commands, you should be able to reproduce the following results:

nDCG@10 BM25
BEIR (v1.0.0): Quora 0.7295
R@100 BM25
BEIR (v1.0.0): Quora 0.9459
R@1000 BM25
BEIR (v1.0.0): Quora 0.9872