riozhcd / trec-research Goto Github PK

• Okapi BM25模型中k1和k3参数的关系
• Zettair 介绍(全文搜索引擎,基于C语言的一个高效倒排索引模式)

zet_trec help information

no topic files given

usage: 'zet_trec index', where index is the name of the index to query
query options:
-f,--file=[topic_file]: add topic_file to list of topic files
-F,--file-list=[file]: add files listed in file to list of topic files
-r,--runid=[run_id]: output run_id as id for this evaluation
-n,--number-results=[num]: number of results to output per query
-t,--title: use title in query
-d,--description: use description in query
-a,--narrative: use narrative in query
--print-queries: print topic queries to stderr
--timing: print total querying time to stderr
(excludes index loading time)
--big-and-fast: use more memory
--dummy: insert dummy results for topics with no results
--non-stop: don't halt on empty topics
--query-stop=[filename]: use filename as stoplist
(or use default if no file give)
--qrels=[filename]: evaluate effectiveness using the given TREC qrels
-h,--help: this message
-v,--version: print version

query metric options:
--anh-impact: evaluate using impact-ordered lists
(must have specified --anh-impact while indexing)
-o,--okapi: use Okapi BM25 metric
-1,--k1=[float]: set Okapi BM25 k1 value
-3,--k3=[float]: set Okapi BM25 k3 value
-b,--b=[float]: set Okapi BM25 b value
-p,--pivoted-cosine=[float]: use pivoted cosine metric, with given pivot
-c,--cosine: use cosine metric
--hawkapi=[float]: use Dave Hawking's metric, with alpha given
--dirichlet=[uint]: use Dirichlet-smoothed LM metric, with mu given

训练模型

接下来的任务是，我们需要在query 451-500上训练出一个模型，使得在501-550查询上能够取得较好的效果