Models: various bag-of-words approaches
This page describes regressions for the TREC 2017 Common Core Track, which uses the New York Times Annotated Corpus. 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 core17
Typical indexing command:
target/appassembler/bin/IndexCollection \
-collection NewYorkTimesCollection \
-input /path/to/nyt \
-index indexes/lucene-index.nyt/ \
-generator DefaultLuceneDocumentGenerator \
-threads 16 -storePositions -storeDocvectors -storeRaw \
>& logs/log.nyt &
The directory /path/to/nyt_corpus/
should be the root directory of the New York Times Annotated Corpus, i.e., ls /path/to/nyt_corpus/
should bring up a bunch of subdirectories, 1987
to 2007
.
For additional details, see explanation of common indexing options.
Topics and qrels are stored in src/main/resources/topics-and-qrels/
, downloaded from NIST:
topics.core17.txt
: topics for the TREC 2017 Common Core Trackqrels.core17.txt
: qrels for the TREC 2017 Common Core Track
After indexing has completed, you should be able to perform retrieval as follows:
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.bm25.topics.core17.txt \
-bm25 &
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.bm25+rm3.topics.core17.txt \
-bm25 -rm3 &
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.bm25+ax.topics.core17.txt \
-bm25 -axiom -axiom.deterministic -rerankCutoff 20 &
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.ql.topics.core17.txt \
-qld &
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.ql+rm3.topics.core17.txt \
-qld -rm3 &
target/appassembler/bin/SearchCollection \
-index indexes/lucene-index.nyt/ \
-topics src/main/resources/topics-and-qrels/topics.core17.txt \
-topicreader Trec \
-output runs/run.nyt.ql+ax.topics.core17.txt \
-qld -axiom -axiom.deterministic -rerankCutoff 20 &
Evaluation can be performed using trec_eval
:
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.bm25.topics.core17.txt
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.bm25+rm3.topics.core17.txt
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.bm25+ax.topics.core17.txt
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.ql.topics.core17.txt
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.ql+rm3.topics.core17.txt
tools/eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt runs/run.nyt.ql+ax.topics.core17.txt
With the above commands, you should be able to reproduce the following results:
MAP | BM25 | +RM3 | +Ax | QL | +RM3 | +Ax |
---|---|---|---|---|---|---|
TREC 2017 Common Core Track Topics | 0.2087 | 0.2823 | 0.2739 | 0.2032 | 0.2606 | 0.2579 |
P30 | BM25 | +RM3 | +Ax | QL | +RM3 | +Ax |
---|---|---|---|---|---|---|
TREC 2017 Common Core Track Topics | 0.4293 | 0.5093 | 0.4940 | 0.4467 | 0.4827 | 0.4893 |