A Method of Chinese Tourism Named Entity Recognition Based on BBLC Model
Leyi Xue, Han Cao, Fan Ye, Yuehua Qin · 2019
In recent years, the booming development of tourism has accumulated a lot of tourism data. The tourism named entity recognition (NER) is of great significance for efficiently obtaining effective information and providing highquality services. However, in terms of data sets, the existing SIGHAN 2006 bakeoff-3 annotation set only contains the people names, location names and organization names, and lacks other types of entity tags. So, this paper embarks from the characteristics of tourism, a certain amount of tourism text was crawled from Ctrip and annotated five types of entities including people name, location name, organization name, time and things with BIO tag rule. On this basis, the BBLC (BERTBiLSTM-CRF) model is constructed which using Google's latest natural language processing model BERT. Compared with BiLSTM-CRF and CRF, the method shows that the method has higher precision, recall rate and F value when performing named entity recognition of tourism data.