Attention Adaptive Chinese Named Entity Recognition Based on Vocabulary Enhancement - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Attention Adaptive Chinese Named Entity Recognition Based on Vocabulary Enhancement

Résumé

To deal with the lack of word information in character vector embedding and the problem of Out-of-Vocabulary in Named Entity Recognition, an attention adaptive chinese named entity recognition (CNER) model based on vocabulary enhancement (ACVE) is proposed. The mechanism of potential information embedding is designed, which acquires word-level potential information by constructing semantic vectors, and the fusion embedding of character information and word-level information realizes the enhancement of semantic features; We also propose an attention mechanism for adaptive distribution, which adaptively adjusts the position of attention by introducing a dynamic scaling factor to obtain the attention distribution suitable for NER tasks. Experiments on a special field dataset with a large number of out-of-vocabulary (OOV) words show that, compared with state-of-the-art methods, our method is more effective and achieves better results.
Fichier sous embargo
Fichier sous embargo
0 7 7
Année Mois Jours
Avant la publication
mercredi 1 janvier 2025
Fichier sous embargo
mercredi 1 janvier 2025
Connectez-vous pour demander l'accès au fichier

Dates et versions

hal-04178713 , version 1 (08-08-2023)

Licence

Paternité

Identifiants

Citer

Ping Zhao, Quansheng Dou, Ping Jiang. Attention Adaptive Chinese Named Entity Recognition Based on Vocabulary Enhancement. 12th International Conference on Intelligent Information Processing (IIP), May 2022, Qingdao, China. pp.399-406, ⟨10.1007/978-3-031-03948-5_32⟩. ⟨hal-04178713⟩
6 Consultations
1 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More