Named Entity Recognition using Multiple Smaller LSTM in Parallel Recurrent Neural Networks

Abstract

In recent years, End-to-End NER with Bidirectional Long-Short-Term Memory (BiLSTM) has attracted increasing attention. However, it remains a great challenge for BiLSTM to have parallel computing, extensive dependencies, and single-function spatial mapping. We propose a deep neural network model based on a parallel computing self-attention mechanism and parallel recurrent neural networks to address these problems. We only use a small amount of BiLSTM to capture the time series of texts and then we use the self-care mechanism, which allows parallel calculations to capture wide-range dependencies.