Tibetan Location Name Recognition Using Tibetan-Chinese Cross-Lingual Word Embeddings
Wei Ma, Kun Zhao · 2019
Location name recognition is an important part of the named entity recognition tasks(NER). For Tibetan Location name recognition, we proposed a method based on Tibetan-Chinese cross-lingual word embedding. We trained Tibetan-Chinese cross-lingual word embedding by using the Tibetan-Chinese bilingual translation word pairs, and proved that the cross-lingual word embedding improves the semantic expression of the monolingual word embedding in semantic similarity experiments. In the BiLSTM-CRF-based Tibetan location name recognition neural network model, we connected the cross-lingual word embedding as the input layer. The experimental results on the Tibetan location name recognition public dataset show that the method of using cross-lingual word embedding effectively improves the performance of Tibetan Location name recognition.