文章目录
0. indri安装以及pyindri安装
indri安装
pyndri安装
1. 构建索引的参数
新建一个manifest.json
文件
<parameters>
<index>Dataset/Robust2004/robust2004_idx(索引地址)</index>
<corpus>
<path>Dataset/Robust2004/Corpus (文档地址)</path>
<class>trectext(文件格式)</class>
</corpus>
<field>
<name>DOCNO(文档编号)</name>
</field>
<field>
<name>TEXT(文档内容)</name>
</field>
</parameters>
2. 建立索引
$ IndriBuildIndex mainfest.json
3. 使用
根据某个query,找到排名靠前的10篇文档
import operator
import pyndri
import sys
index_path='../../Dataset/Robust2004/robust2004_idx'
with pyndri.open(index_path) as index:
results = index.query('hello world', results_requested=10)
# print(results)
for int_document_id, score in results:
ext_document_id, document_terms = index.document(int_document_id)
# print(ext_document_id)
# print(document_terms)
print('Document {ext_document_id} retrieved with score {score}.'.format(
ext_document_id=ext_document_id, score=score))
>>>
Document LA081890-0076 retrieved with score -6.731433614635001.
Document LA102090-0133 retrieved with score -6.734136943420365.
Document FT942-1058 retrieved with score -6.904603734669878.
Document FBIS4-19258 retrieved with score -6.948269332125547.
Document LA093089-0055 retrieved with score -6.969842877219223.
print(results)
>>>
((485281, -6.731433614635001),
(234645, -6.734136943420365),
(96799, -6.904603734669878),
(478636, -6.948269332125547),
(285324, -6.969842877219223))
results
中的5
个元组分别对应检索返回的5
篇文档
每个元组的第一个元素代表的是文档id
,第二个元素代表的是得分,得分是负的,是因为这个分数是取了log之后的结果
ext_document_id, document_terms = index.document(int_document_id)
print(ext_document_id) # 文档真正的编号
>>>
LA081890-0076
print(document_terms) # 文档中的词,词的id
>>>
(519720, 14369, 673, 318, 202, 725, 4711, 197,...,2, 480, 2, 1, 8696)
计算某个query和指定文档之间的分数
import operator
import pyndri
import sys
index_path='../../Dataset/Robust2004/robust2004_idx'
with pyndri.open(index_path) as index:
# 单个query与指定文档之间的分数
results = index.query(
'hello world',
document_set=map(
operator.itemgetter(1),
index.document_ids(['LA081890-0076', 'LA102090-0133'])))
# print(results)
for int_document_id, score in results:
ext_document_id, _ = index.document(int_document_id)
print('Document {ext_document_id} '
'retrieved with score {score}.'.format(
ext_document_id=ext_document_id,
score=score))
>>>
Document LA081890-0076 retrieved with score -6.731433614635001.
Document LA102090-0133 retrieved with score -6.734136943420365.
>>> print(result)
((485281, -6.731433614635001), (234645, -6.734136943420365))
index.document_ids(['LA081890-0076', 'LA102090-0133'])
返回的是2
个元组,每个元组的第一个元素是文档编号
,第二个元素是文档id
operator.itemgetter(1)
代表的是取出第一维的数据,也就是文档编号
>>> index.document_ids(['LA081890-0076', 'LA102090-0133'])
(('LA081890-0076', 485281), ('LA102090-0133', 234645))