Research On Kazakh Syntactic Parsing Auxiliary Feature Extraction

The processing of Kazakh language has now completed basic phrase automatic recognition,fixed phrase extraction,and part-of-speech tagging.This paper focuses on the parsing of the phrase structure in Kazakh language.This article mainly adopts the syntactic parsing method based on shift-reduce,uses perceptron algorithm and beamsearch algorithm to train and decode in the process of syntactic parsing,and adds neural network to add auxiliary features for the decoding process to help generate Kazakh syntactic structure more efficiently.The main research work of this paper is as follows:1.In the Kazakh syntactic parsing process,the method based on shift-reduce is used as the overall framework,and the syntactic tree of Kazakh phrase structure is finally obtained through the shift-reduce action of each step.2.In the Kazakh syntactic parsing training stage,the perceptron algorithm is used to train the sentences of the standard Kazakh sentence structure.The traditional syntactic parsing mainly relies on the language rule base,and the perceptron algorithm training is a continuous adjustment parameter to continuously learn the Kazakh sentence.It initializes the parameter vector to zero and updates the parameters by decoding the training samples.3.In the Kazakh syntactic parsing decoding stage,the beam-search algorithm is used to decode.In the process of parsing the sentence,the size of the search space is controlled by giving up some small probability nodes with relatively small weights to obtain a syntactic tree with a large probability.It not only controls the search space occupied by the decoding process,but also obtains the most suitable syntactic tree through statistical methods.4.In the process of decoding,auxiliary features are added through neural networks.In the process of parsing sentences,a bi-directional LSTM model is constructed in this paper.It is used to extract the information of the structure of each word in a sentence to predict each word in the syntactic tree.Syntactic components are then passed on to the syntactic parsing process as an auxiliary look-ahead feature to help the decoder generate the Kazakh syntactic tree.Experiments show that the Kazakh syntactic structure can be analyzed when using the syntactic parsing method based on shift-reduce,and the neural network is used to extract some look-ahead features to assist decoding in the syntactic parsing process.