Emotional Analysis Of Microblog Text Research

With the advancement and development of the Internet,there are a lot of data spread on the internet every day.These data contains great value.Among them,microblog as a representative of social networks,attracting a large number of users in which to discuss hot spots,express their views.Fast and accurate analysis of microblog user's emotions,contains a huge commercial and political value.The object of the text study is the emotional analysis for the microblog text.Firstly,the first two kinds of emotional analysis schemes are analyzed,which are the methods of emotional dictionary and the method of machine learning.This paper presents a fine-grained emotional analysis method that combines emotional dictionary and machine learning.The emotional dictionary as a sentence-level emotional analysis tool,in the construction of emotional dictionary method,based on the Dalian Institute of emotional ontology library,the network of new words,word combinations,facial expressions and repetitive tone words added to the emotional dictionary.And taking into account the influence of degree adverbs and negative words on emotional expression,we also generate negative vocabularies and adverbs.Experiments show that the emotional dictionary extended in this paper has better effect on the emotional analysis of microblog text than the emotional ontology library.Based on the advantages of SVM in the small sample training set,the text adopts SVM as the classifier of machine learning.This paper discusses the application of SVM in multi-class classification.After analyzing and comparing several other multi-class SVM classifiers.This paper improves and optimizes based on the binary tree SVM multi-classification algorithm.Respectively,in the training phase and classification stage are the corresponding optimization,by reducing the cumulative error and reduce the invalid calculation to improve the classification accuracy and classification efficiency.In order to consider the rules between the emotional attributes of the sentence,this paper proposes a kind of sequence rule,which labels the sentences as emotional labels,sets the emotional categories and conjunctions to train and classifies the sequence rules.And treat it as one of the characteristics of classification,together with the sentence of emotional word characteristics,sentence structure,symbolic characteristics.To take full account of their impact on the emotional category of text.