Skip to content

Latest commit

 

History

History
79 lines (63 loc) · 4.81 KB

regressions-ciral-v1.0-yo-en.md

File metadata and controls

79 lines (63 loc) · 4.81 KB
Raw

Anserini Regressions: CIRAL (v1.0) — Yoruba (English Translation)

This page documents BM25 regression experiments for CIRAL (v1.0) — Yoruba with with document translations. To be clear, the queries are in English and the corpus is in English (translated with NLLB 1.3B).

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 ciral-v1.0-yo-en

Indexing

Typical indexing command:

target/appassembler/bin/IndexCollection \
  -collection MrTyDiCollection \
  -input /path/to/ciral-yoruba-english \
  -generator DefaultLuceneDocumentGenerator \
  -index indexes/lucene-index.ciral-v1.0-yo-en/ \
  -threads 16 -storePositions -storeDocvectors -storeRaw \
  >& logs/log.ciral-yoruba-english &

See this page for more details about the CIRAL corpus. For additional details, see explanation of common indexing options.

Retrieval

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

target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.ciral-v1.0-yo-en/ \
  -topics tools/topics-and-qrels/topics.ciral-v1.0-yo-test-a.tsv \
  -topicReader TsvInt \
  -output runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt \
  -bm25 -hits 1000 &
target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.ciral-v1.0-yo-en/ \
  -topics tools/topics-and-qrels/topics.ciral-v1.0-yo-test-a.tsv \
  -topicReader TsvInt \
  -output runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt \
  -bm25 -hits 1000 &
target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.ciral-v1.0-yo-en/ \
  -topics tools/topics-and-qrels/topics.ciral-v1.0-yo-test-b.tsv \
  -topicReader TsvInt \
  -output runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-b.txt \
  -bm25 -hits 1000 &

Evaluation can be performed using trec_eval:

target/appassembler/bin/trec_eval -c -m ndcg_cut.20 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-a.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt
target/appassembler/bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-a.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt
target/appassembler/bin/trec_eval -c -m ndcg_cut.20 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-a-pools.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt
target/appassembler/bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-a-pools.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-a.txt
target/appassembler/bin/trec_eval -c -m ndcg_cut.20 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-b.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-b.txt
target/appassembler/bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.ciral-v1.0-yo-test-b.tsv runs/run.ciral-yoruba-english.bm25-default.topics.ciral-v1.0-yo-test-b.txt

Effectiveness

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

nDCG@20 BM25 (default)
CIRAL Yoruba: Test Set A (Shallow Judgements) 0.4265
CIRAL Yoruba: Test Set A (Pools) 0.4451
CIRAL Yoruba: Test Set B 0.3700
R@100 BM25 (default)
CIRAL Yoruba: Test Set A (Shallow Judgements) 0.7832
CIRAL Yoruba: Test Set A (Pools) 0.7199
CIRAL Yoruba: Test Set B 0.7348