Skip to content

Latest commit

 

History

History
68 lines (49 loc) · 4 KB

regressions-beir-v1.0.0-hotpotqa-unicoil-noexp.md

File metadata and controls

68 lines (49 loc) · 4 KB
Raw

Anserini Regressions: BEIR (v1.0.0) — HotpotQA

Model: uniCOIL (without any expansions)

This page describes regression experiments, integrated into Anserini's regression testing framework, using uniCOIL (without any expansions) on BEIR (v1.0.0) — HotpotQA. The uniCOIL model is described in the following paper:

Jimmy Lin and Xueguang Ma. A Few Brief Notes on DeepImpact, COIL, and a Conceptual Framework for Information Retrieval Techniques. arXiv:2106.14807.

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-hotpotqa-unicoil-noexp

Indexing

Typical indexing command:

target/appassembler/bin/IndexCollection \
  -collection JsonVectorCollection \
  -input /path/to/beir-v1.0.0-hotpotqa-unicoil-noexp \
  -index indexes/lucene-index.beir-v1.0.0-hotpotqa-unicoil-noexp/ \
  -generator DefaultLuceneDocumentGenerator \
  -threads 16 -impact -pretokenized \
  >& logs/log.beir-v1.0.0-hotpotqa-unicoil-noexp &

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:

target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.beir-v1.0.0-hotpotqa-unicoil-noexp/ \
  -topics tools/topics-and-qrels/topics.beir-v1.0.0-hotpotqa.test.unicoil-noexp.tsv.gz \
  -topicreader TsvString \
  -output runs/run.beir-v1.0.0-hotpotqa-unicoil-noexp.unicoil-noexp.topics.beir-v1.0.0-hotpotqa.test.unicoil-noexp.txt \
  -impact -pretokenized -removeQuery -hits 1000 &

Evaluation can be performed using trec_eval:

tools/eval/trec_eval.9.0.4/trec_eval -c -m ndcg_cut.10 tools/topics-and-qrels/qrels.beir-v1.0.0-hotpotqa.test.txt runs/run.beir-v1.0.0-hotpotqa-unicoil-noexp.unicoil-noexp.topics.beir-v1.0.0-hotpotqa.test.unicoil-noexp.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.beir-v1.0.0-hotpotqa.test.txt runs/run.beir-v1.0.0-hotpotqa-unicoil-noexp.unicoil-noexp.topics.beir-v1.0.0-hotpotqa.test.unicoil-noexp.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.beir-v1.0.0-hotpotqa.test.txt runs/run.beir-v1.0.0-hotpotqa-unicoil-noexp.unicoil-noexp.topics.beir-v1.0.0-hotpotqa.test.unicoil-noexp.txt

Effectiveness

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

nDCG@10 uniCOIL no expansion
BEIR (v1.0.0): HotpotQA 0.6671
R@100 uniCOIL no expansion
BEIR (v1.0.0): HotpotQA 0.7978
R@1000 uniCOIL no expansion
BEIR (v1.0.0): HotpotQA 0.8722