Skip to content

Latest commit

 

History

History

Weibo

readme for Weibo NER dataset

Task

Named Entity Recognition

Description

Tags: PER(人名), LOC(地点名), GPE(行政区名), ORG(机构名)

Label Tag Meaning
PER PER.NAM 名字(张三)
PER.NOM 代称、类别名(穷人)
LOC LOC.NAM 特指名称(紫玉山庄)
LOC.NOM 泛称(大峡谷、宾馆)
GPE GPE.NAM 行政区的名称(北京)
ORG ORG.NAM 特定机构名称(通惠医院)
ORG.NOM 泛指名称、统称(文艺公司)

Tag Strategy:BIO
Split:
'\t' in raw data (北\tB-LOC)
'space'in transformed data (北 B-LOC)
Data Size:
Train data set ( weiboNER.conll.train ):

句数 字符数 PER.NAM数 PER.NOM数 LOC.NAM数 LOC.NOM数 GPE.NAM数 ORG.NAM数 ORG.NOM数
1350 73778 574 766 56 51 205 183 42

Dev data set ( weiboNER.conll.dev ):

句数 字符数 PER.NAM数 PER.NOM数 LOC.NAM数 LOC.NOM数 GPE.NAM数 ORG.NAM数 ORG.NOM数
270 14509 90 208 6 6 26 47 5

Test data set ( weiboNER.conll.test )

句数 字符数 PER.NAM数 PER.NOM数 LOC.NAM数 LOC.NOM数 GPE.NAM数 ORG.NAM数 ORG.NOM数
270 14842 111 170 19 9 47 39 17

Note:
the raw data contains segmentation information, like:

她0	O
和0	O
现0	B-PER.NOM
任1	I-PER.NOM
男0	B-PER.NOM
友1	I-PER.NOM
交0	O
往1	O
时0	O

the sample conll file should not contain segmentation, like:

她 O
和 O
现 B-PER.NOM
任 I-PER.NOM
男 B-PER.NOM
友 I-PER.NOM
交 O
往 O
时 O

a script was provided to transform the raw data, in which segmentation was deleted, '\t' was transformed to 'space', you can use following command.

python3 transform_data.py

Reference:
Named Entity Recognition for Chinese Social Media with Jointly Trained Embeddings
Improving Named Entity Recognition for Chinese Social Media with Word Segmentation Representation Learning
https://github.com/hltcoe/golden-horse