Skip to content

Latest commit

 

History

History
74 lines (54 loc) · 4.65 KB

regressions-msmarco-passage-hgf-wp.md

File metadata and controls

74 lines (54 loc) · 4.65 KB
Raw

Anserini Regressions: MS MARCO Passage Ranking

Models: bag-of-words approaches with WordPiece tokenization provided by HuggingFace Tokenizer Integration

This page documents regression experiments on the MS MARCO passage ranking task, which is integrated into Anserini's regression testing framework. Here we are using WordPiece tokenization (i.e., from BERT) with the following tokenizer from HuggingFace bert-base-uncased: .

In general, effectiveness is lower than with "standard" Lucene tokenization for two reasons: (1) we're losing stemming, and (2) some terms are chopped into less meaningful subwords.

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 msmarco-passage-hgf-wp

Indexing

Typical indexing command:

target/appassembler/bin/IndexCollection \
  -collection JsonCollection \
  -input /path/to/msmarco-wp \
  -index indexes/lucene-index.msmarco-passage-hgf-wp/ \
  -generator DefaultLuceneDocumentGenerator \
  -threads 9 -storePositions -storeDocvectors -storeRaw -analyzeWithHuggingFaceTokenizer bert-base-uncased \
  >& logs/log.msmarco-wp &

The directory /path/to/msmarco-passage-wp/ should be a directory containing the corpus in Anserini's jsonl format.

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. The regression experiments here evaluate on the 6980 dev set questions; see this page for more details.

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

target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.msmarco-passage-hgf-wp/ \
  -topics tools/topics-and-qrels/topics.msmarco-passage.dev-subset.txt \
  -topicreader TsvInt \
  -output runs/run.msmarco-wp.bm25-default.topics.msmarco-passage.dev-subset.txt \
  -bm25  -analyzeWithHuggingFaceTokenizer  bert-base-uncased &

Evaluation can be performed using trec_eval:

tools/eval/trec_eval.9.0.4/trec_eval -c -m map tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-wp.bm25-default.topics.msmarco-passage.dev-subset.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -M 10 -m recip_rank tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-wp.bm25-default.topics.msmarco-passage.dev-subset.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-wp.bm25-default.topics.msmarco-passage.dev-subset.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-wp.bm25-default.topics.msmarco-passage.dev-subset.txt

Effectiveness

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

AP@1000 BM25 (default)
MS MARCO Passage: Dev 0.1836
RR@10 BM25 (default)
MS MARCO Passage: Dev 0.1752
R@100 BM25 (default)
MS MARCO Passage: Dev 0.6231
R@1000 BM25 (default)
MS MARCO Passage: Dev 0.8263