Research On Mongolian Syntactic Parsing Based On Deep Learning

Mongolian syntactic parsing is one of the important research contents in the Mongolian information processing.It is an important foundation for the research including text processing,semantic research,machine translation,automatic question answering system research and automatic information extraction.In recent years,the deep learning and neural network based method,which is a research hotspot,has made important research progress in the fields of speech recognition,image processing,handwriting recognition,and medicine.Combining it with natural language processing tasks is a new approach to natural language processing research.Therefore,based on the deep learning and neural network,this paper makes a initial research on the Mongolian syntactic parsing.In this paper,three Mongolian syntactic parsing models are constructed.The first is a Mongolian dependency parsing model based on recurrent neural networks.This model uses a dependency parsing model based on a feed-forward neural network as a feature extractor.The extracted features are used as input to the recurrent neural network,and the Mongolian dependency parsing model based on recurrent neural network is trained.The Mongolian sentence is an ordered,continuous string of words,and the commonly used recurrent neural network has the advantage of time series.Therefore,the training of the Mongolian dependency parsing model based on recurrent neural network has achieved certain effects.Secondly,the Mongolian dependency parsing model based on long short-term memory neural networks.The construction of this model is basically consistent with the Mongolian dependency parsing model based on recurrent neural network.The difference is that the long short-term memory neural network is an extension of the recurrent neural network.The long short-term memory neural network solves the problem of the recurrent neural network gradient disappearing by introducing the concept of memory cell and control gate.The network can also remember the relationship between two words with long time intervals,so that compared with the Mongolian dependency parsing model based on recurrent neural network,the accuracy of Mongolian dependency parsing model based on long short-term memory neural network has a certain increase.Finally,it is based on the end-to-end Mongolian phrase structure syntax analysis model.This model uses a sequence-to-sequence model consisting of an encoder and a decoder.The Mongolian sentence is used as the source language,and the phrase structure syntax tree is used as the target language to train the end-to-end Mongolian phrase structure syntax analysis model,and some results have been achieved.In the conclusion part,summarizes the full text of the paper and makes a brief discussion on the future research work.