Anserini: Regressions for AQUAINT (Robust05)
Typical indexing command:
nohup sh target/appassembler/bin/IndexCollection -collection TrecCollection \
-generator JsoupGenerator -threads 16 -input /path/to/robust05 -index \
lucene-index.robust05.pos+docvectors+rawdocs -storePositions -storeDocvectors \
-storeRawDocs >& log.robust05.pos+docvectors+rawdocs &
The directory /path/to/aquaint/ should be the root directory of AQUAINT collection; under subdirectory disk1/ there should be NYT/ and under subdirectory disk2/ there should be APW/ and XIE/.
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.robust05.txt: Hard Topics of Robust04 (TREC 2005 Robust Track)qrels.robust2005.txt: qrels (TREC 2005 Robust Track)
After indexing has completed, you should be able to perform retrieval as follows:
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.bm25.topics.robust05.txt -bm25 &
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.bm25+rm3.topics.robust05.txt -bm25 -rm3 &
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.bm25+ax.topics.robust05.txt -bm25 -axiom -rerankCutoff 20 -axiom.deterministic &
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.ql.topics.robust05.txt -ql &
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.ql+rm3.topics.robust05.txt -ql -rm3 &
nohup target/appassembler/bin/SearchCollection -topicreader Trec -index lucene-index.robust05.pos+docvectors+rawdocs -topics src/main/resources/topics-and-qrels/topics.robust05.txt -output run.robust05.ql+ax.topics.robust05.txt -ql -axiom -rerankCutoff 20 -axiom.deterministic &
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.robust2005.txt run.robust05.bm25.topics.robust05.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2005.txt run.robust05.bm25+rm3.topics.robust05.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2005.txt run.robust05.bm25+ax.topics.robust05.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2005.txt run.robust05.ql.topics.robust05.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2005.txt run.robust05.ql+rm3.topics.robust05.txt
eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2005.txt run.robust05.ql+ax.topics.robust05.txt
With the above commands, you should be able to replicate the following results:
| MAP | BM25 | +RM3 | +Ax | QL | +RM3 | +Ax |
|---|---|---|---|---|---|---|
| TREC 2005 Robust Track Topics | 0.2032 | 0.2602 | 0.2587 | 0.2028 | 0.2491 | 0.2476 |
| P30 | BM25 | +RM3 | +Ax | QL | +RM3 | +Ax |
|---|---|---|---|---|---|---|
| TREC 2005 Robust Track Topics | 0.3693 | 0.4187 | 0.4120 | 0.3653 | 0.4067 | 0.4113 |