Word Embedding에 PCA를 적용한 개체명 인식 모델을 위한 효율적인 학습방법 연구

Abstract

The Bidirectional LSTM CRF model used for Named Entity Recognition takes much time to train NamedEntity. The hyper-parameters of Word Embedding used as input data in this model affect performance andtraining time. However, there is very little research on the number of dimensions, which is one of the parametersof Word Embedding. In this paper, we obtain proper number of 4-Word Embeddings (fastText, GloVe, skipgram,CBOW) considering performance and training time in Bidirectional LSTM CRF which can input largeamount of data. Next, apply the PCA to the word vector in Word Embedding to reduce the dimension to smalldimensional (10 dimensions) intervals. Therefore, applying PCA to conventional Word Embedding and trainingWord Embedding with small dimensional intervals shows that the model can be trained by maintaining orimproving performance based on stable training time in fewer dimensions than conventional Word Embedding