Research on Coreference Resolution Based on Conditional Random Fields

Abstract

In view of the phenomenon of noun coreference in Chinese, This paper proposes a deep learning mechanism based on Conditional Random Field (CRF) to study coreference resolution based on deep semantic information representation. The text is input into the vector of Biditive Encoder Representations from Transformers model. The self-attention mechanism is used to mine the hidden features at the context semantic level. Through the reasoning ability of CRF, the complex features are used for reasoning training, and the training results are scored and classified by softmax to complete the anaphora resolution task. The experimental results show that the performance of coreference resolution can be effectively improved by making full use of text feature representation.