A Study On Sentence Uncertainty Identification And Classification

Uncertainty can generally be interpreted as a lack of information in linguistics.People often use uncertain expressions when they lack some relevant information,such as the inference in academic papers,the gossip on the Internet and etc.As the era of big data comes,researchers pay more attention to the study of uncertainty identification and classification to meet the need of text processing.In early researches,uncertainty identification was used to extract information accurately.Researchers built uncertain corpora based on academic papers and Wikipedia,and mainly use classification methods based on cues.But the cues did not always work.What's more,there was few studies on uncertainty classification.To solve these problems,I built a corpus and designed a model without cues.Using social media text,I built a Chinese Social Media Uncertain Corpus.Not only the uncertainty was labeled,but also the type of uncertainty was labeled in the corpus.A deep learning model was applied to judge uncertainty.With a recurrent neural network to understand sentence semantics,attention mechanisms to capture important uncertain features and a convolutional neural network to classified sentence,our model performed well.Comparing with some existing methods in experiments,it proved that my model could obtain the best results in most cases.In conclusion,the purpose of this paper is to promote the study on sentence uncertainty.A Chinese social media uncertainty corpus is construct,and I proposed a new deep learning model,which has excellent performance in the uncertainty identification and classification.The research results will be helpful to Natural Language Processing problems,so as to promote the related researches.