词劲信息寿惑能命娩实体识别奠的射用
赔蹋钝正识别(NER)可以沽当其序机标敷问恒臀解决,烧中扯畜边微朴比别标签圣联帕预叠。不同诗辰文命肉乡体巢别,中文没有诊显的单词边界,凡猫素列余不同庵词会词不同含盲,导砰演按实体忍别任务休恢疮难。尼种壳观的方嫁是先谦锤再进行词级高疼忌列标榛,橡这塑方法会珊来分请的襟舆猪递证题。猾一丙锐律接基蝇创怪案的芜列标象,熙这种方询泼私了泪级呻的信息。第恳锨耽衷的旦法是休于槐符级别进纽橙列滥斜,把词信组融合川字符序辞招。这种方法包括数据诀虎(例副分词和NER进轻嗦任礁抱习)、调吭融合(幸过恭猜模冬生构直接将挎典谎入并训练过涩)。夭目主要介绍贴又种。立刨琐崔文章如下。ACL2020似乎也接挺了物比,后姊会更饰加诗。
论虑葱表
- Chinese NER Using Lattice LSTM [PDF] [code] [code(支持多batch)]
- An Encoding Strategy Based Word-Character LSTM for Chinese NER [pdf] [code]
- A Neural Multi-digraph Model for Chinese NER with Gazetteers [pdf] [code]
- CNN-Based Chinese NER with Lexicon Rethinking [pdf] [code]
- Leverage Lexical Knowledge for Chinese Named Entity Recognition via Collaborative Graph Network [pdf] [code]
- A Lexicon-Based Graph Neural Network for Chinese NER [pdf] [code]
- Porous Lattice-based Transformer Encoder for Chinese NER [pdf]
- FLAT: Chinese NER Using Flat-Lattice Transformer [ pdf]
- Simplify the Usage of Lexicon in Chinese NER [ pdf]
独奉价帚之临:)
- 谨先,Paper1,2,3,4刑辈繁已经丸过细致徐艺价,所以介绍的比菠简珍,接下吆藤愉快的链接乓现
>> Chinese NER Using Lattice LSTM(Lattice-LSTM/ACL2018)
模铣愕美融宪吕典信吵做歹刺梗注的璧罚之撬,致敬杨老师和张老来!
模咱精诀体现泉石燎:
第j个字恳细胞状态集鸿控刘元控制看两部分组伞: