Anserini: Regressions for the New York Times (Core17)
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.
Typical indexing command:
nohup sh target/appassembler/bin/IndexCollection -collection NewYorkTimesCollection -input /path/to/core17 \
-index indexes/lucene-index.core17.pos+docvectors+raw -generator DefaultLuceneDocumentGenerator -threads 16 \
-storePositions -storeDocvectors -storeRaw >& logs/log.core17.pos+docvectors+rawdocs &
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:
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-bm25 -output run.core17.bm25.topics.core17.txt &
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-bm25 -rm3 -output run.core17.bm25+rm3.topics.core17.txt &
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-bm25 -axiom -axiom.deterministic -rerankCutoff 20 -output run.core17.bm25+ax.topics.core17.txt &
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-qld -output run.core17.ql.topics.core17.txt &
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-qld -rm3 -output run.core17.ql+rm3.topics.core17.txt &
nohup target/appassembler/bin/SearchCollection -index indexes/lucene-index.core17.pos+docvectors+raw \
-topicreader Trec -topics src/main/resources/topics-and-qrels/topics.core17.txt \
-qld -axiom -axiom.deterministic -rerankCutoff 20 -output run.core17.ql+ax.topics.core17.txt &
Evaluation can be performed using trec_eval
:
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.bm25.topics.core17.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.bm25+rm3.topics.core17.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.bm25+ax.topics.core17.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.ql.topics.core17.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.ql+rm3.topics.core17.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.core17.txt run.core17.ql+ax.topics.core17.txt
With the above commands, you should be able to replicate 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 |