Skip to content

Latest commit

 

History

History
64 lines (48 loc) · 2.94 KB

regressions-hc4-v1.0-zh.md

File metadata and controls

64 lines (48 loc) · 2.94 KB
Raw

Anserini Regressions: HC4 (v1.0) — Chinese

This page documents BM25 regression experiments for HC4 (v1.0) — Chinese.

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 hc4-v1.0-zh

Indexing

Typical indexing command:

target/appassembler/bin/IndexCollection \
  -collection NeuClirCollection \
  -input /path/to/hc4-v1.0-zh \
  -index indexes/lucene-index.hc4-v1.0-chinese/ \
  -generator DefaultLuceneDocumentGenerator \
  -threads 8 -storePositions -storeDocvectors -storeRaw -language zh \
  >& logs/log.hc4-v1.0-zh &

See this page for more details about the HC4 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.hc4-v1.0-chinese/ \
  -topics src/main/resources/topics-and-qrels/topics.hc4-v1.0-zh.dev.title.tsv.gz \
  -topicreader TsvInt \
  -output runs/run.hc4-v1.0-zh.bm25.topics.hc4-v1.0-zh.dev.title.txt \
  -bm25 -hits 100 -language zh &
target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.hc4-v1.0-chinese/ \
  -topics src/main/resources/topics-and-qrels/topics.hc4-v1.0-zh.dev.desc.tsv.gz \
  -topicreader TsvInt \
  -output runs/run.hc4-v1.0-zh.bm25.topics.hc4-v1.0-zh.dev.desc.txt \
  -bm25 -hits 100 -language zh &

Evaluation can be performed using trec_eval:

tools/eval/trec_eval.9.0.4/trec_eval -c -M 100 -m map src/main/resources/topics-and-qrels/qrels.hc4-v1.0-zh.dev.txt runs/run.hc4-v1.0-zh.bm25.topics.hc4-v1.0-zh.dev.title.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -M 100 -m map src/main/resources/topics-and-qrels/qrels.hc4-v1.0-zh.dev.txt runs/run.hc4-v1.0-zh.bm25.topics.hc4-v1.0-zh.dev.desc.txt

Effectiveness

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

MAP BM25
HC4 (Chinese): dev-topic title 0.2914
HC4 (Chinese): dev-topic description 0.1983