Extraction And Analysis Of Event Information In Uyghur Language Based On Deep Learning

With the popularity of Internet and explosive growth of network information,people are getting more and more attention from the internet.How to extract relevant information efficiently and accurately becomes very challenging.Event extraction is an important research direction in the field of information extraction,the events as the basic unit of information representation and organization methods,the unstructured text contains event information in a structured or semi-structured form from different sources extracted and presented in different levels and size.Extraction and analysis of the Uighur language event aims to use deep learning technology from unstructured data in the event of Uighur language information extraction,automatic text summarization,question answering system is based on the basis of information organization technology of the event.With the rapid development of communications in Xinjiang,a large number of websites and communication platforms based on ethnic languages have been built up,which provide a large number of corpus resources for the extraction and analysis of Uyghur event information.Event information extraction and analysis include event recognition and event element identification.Event recognition aims to extract event triggered words and classify their corresponding events.In the event of event recognition sentence context latent semantic information mining is not sufficient,identify the problem of weak stability,combined with the analysis of Uyghur language characteristics,extracting six features of the Uighur language event contains the block.In order to improve Uyghur event feature representation,Word Embedding,which is rich in vocabulary,semantic and context location,is introduced into feature set.The depth learning model is used to abstract the learning ability of feature abstraction and the capture of the sequence of abstract meaning in the event,and the classifier is trained to perform the event recognition task.The experimental results show that the introduction of Word Embedding feature improves the performance of model recognition,and achieves better recall and accuracy as well as F value with the fusion of six feature blocks.In the part of event element identification,this paper uses deep learning technology to translate it into classification problem,and identifies the real event element categories from the identified event categories.And focus on the training of model classifier and the selection of features,and the discovery of hidden features.The lack of excessive manual intervention and lack of relevant domain knowledge results in more objective recognition results.Experimental results show that the depth learning technology can effectively improve the training instance imbalance and data sparseness in Uyghur language event recognition task,and enhance the stability of system identification.