Skip to content

Latest commit

 

History

History
136 lines (99 loc) · 9.54 KB

regressions-msmarco-v2-doc-segmented-unicoil-0shot.md

File metadata and controls

136 lines (99 loc) · 9.54 KB
Raw

Anserini Regressions: MS MARCO (V2) Document Ranking

Model: uniCOIL (with doc2query-T5 expansions) zero-shot on segmented documents (segment-only encoding) - Deprecated, see below

This page describes regression experiments for document ranking on the segmented version of the MS MARCO (V2) document corpus using the dev queries, which is integrated into Anserini's regression testing framework. Here, we cover experiments with the uniCOIL model trained on the MS MARCO V1 passage ranking test collection, applied in a zero-shot manner, with doc2query-T5 expansion.

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.

NOTE: As an important detail, there is the question of what text we feed into the encoder to generate document representations. Initially, we fed only the segment text, but later we realized that prepending the title of the document improves effectiveness. This regression captures segment-only encoding and is kept around primarily for archival purposes; you probably don't want to use this one unless you're running ablation experiments. The version that uses title/segment encoding can be found here.

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-v2-doc-segmented-unicoil-0shot

We make available a version of the MS MARCO document corpus that has already been processed with uniCOIL (per above), i.e., we have applied doc2query-T5 expansions, performed model inference on every document, and stored the output sparse vectors. Thus, no neural inference is involved.

From any machine, the following command will download the corpus and perform the complete regression, end to end:

python src/main/python/run_regression.py --download --index --verify --search --regression msmarco-v2-doc-segmented-unicoil-0shot

The run_regression.py script automates the following steps, but if you want to perform each step manually, simply copy/paste from the commands below and you'll obtain the same regression results.

Corpus Download

Download, unpack, and prepare the corpus:

# Download
wget https://rgw.cs.uwaterloo.ca/JIMMYLIN-bucket0/data/msmarco_v2_doc_segmented_unicoil_0shot.tar -P collections/

# Unpack
tar -xvf collections/msmarco_v2_doc_segmented_unicoil_0shot.tar -C collections/

# Rename (indexer is expecting corpus under a slightly different name)
mv collections/msmarco_v2_doc_segmented_unicoil_0shot collections/msmarco-v2-doc-segmented-unicoil-0shot

To confirm, msmarco_v2_doc_segmented_unicoil_0shot.tar is 62 GB and has an MD5 checksum of 889db095113cc4fe152382ccff73304a. With the corpus downloaded, the following command will perform the remaining steps below:

python src/main/python/run_regression.py --index --verify --search --regression msmarco-v2-doc-segmented-unicoil-0shot \
  --corpus-path collections/msmarco-v2-doc-segmented-unicoil-0shot

Indexing

Sample indexing command:

target/appassembler/bin/IndexCollection \
  -collection JsonVectorCollection \
  -input /path/to/msmarco-v2-doc-segmented-unicoil-0shot \
  -index indexes/lucene-index.msmarco-v2-doc-segmented-unicoil-0shot/ \
  -generator DefaultLuceneDocumentGenerator \
  -threads 18 -impact -pretokenized \
  >& logs/log.msmarco-v2-doc-segmented-unicoil-0shot &

The path /path/to/msmarco-v2-doc-segmented-unicoil-0shot/ should point to the corpus downloaded above.

The important indexing options to note here are -impact -pretokenized: the first tells Anserini not to encode BM25 doclengths into Lucene's norms (which is the default) and the second option says not to apply any additional tokenization on the uniCOIL tokens. Upon completion, we should have an index with 124,131,414 documents.

For additional details, see explanation of common indexing options.

Retrieval

Topics and qrels are stored in src/main/resources/topics-and-qrels/. These regression experiments use the dev queries and the dev2 queries.

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

target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.msmarco-v2-doc-segmented-unicoil-0shot/ \
  -topics src/main/resources/topics-and-qrels/topics.msmarco-v2-doc.dev.unicoil.0shot.tsv.gz \
  -topicreader TsvInt \
  -output runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev.unicoil.0shot.txt \
  -impact -pretokenized -hits 10000 -selectMaxPassage -selectMaxPassage.delimiter "#" -selectMaxPassage.hits 1000 &
target/appassembler/bin/SearchCollection \
  -index indexes/lucene-index.msmarco-v2-doc-segmented-unicoil-0shot/ \
  -topics src/main/resources/topics-and-qrels/topics.msmarco-v2-doc.dev2.unicoil.0shot.tsv.gz \
  -topicreader TsvInt \
  -output runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev2.unicoil.0shot.txt \
  -impact -pretokenized -hits 10000 -selectMaxPassage -selectMaxPassage.delimiter "#" -selectMaxPassage.hits 1000 &

Evaluation can be performed using trec_eval:

tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.100 src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev.unicoil.0shot.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.1000 src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev.unicoil.0shot.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -M 100 -m map -c -M 100 -m recip_rank src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev.unicoil.0shot.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.100 src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev2.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev2.unicoil.0shot.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -m recall.1000 src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev2.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev2.unicoil.0shot.txt
tools/eval/trec_eval.9.0.4/trec_eval -c -M 100 -m map -c -M 100 -m recip_rank src/main/resources/topics-and-qrels/qrels.msmarco-v2-doc.dev2.txt runs/run.msmarco-v2-doc-segmented-unicoil-0shot.unicoil-0shot.topics.msmarco-v2-doc.dev2.unicoil.0shot.txt

Effectiveness

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

MAP@100 uniCOIL (with doc2query-T5) zero-shot
MS MARCO V2 Doc: Dev 0.2218
MS MARCO V2 Doc: Dev2 0.2270
MRR@100 uniCOIL (with doc2query-T5) zero-shot
MS MARCO V2 Doc: Dev 0.2243
MS MARCO V2 Doc: Dev2 0.2291
R@100 uniCOIL (with doc2query-T5) zero-shot
MS MARCO V2 Doc: Dev 0.7551
MS MARCO V2 Doc: Dev2 0.7550
R@1000 uniCOIL (with doc2query-T5) zero-shot
MS MARCO V2 Doc: Dev 0.9056
MS MARCO V2 Doc: Dev2 0.9097

Reproduction Log*

To add to this reproduction log, modify this template and run bin/build.sh to rebuild the documentation.