Font Size: a A A

Research On Tencent Weibo Emotional Tendency Based On Emoji

Posted on:2018-01-05Degree:MasterType:Thesis
Country:ChinaCandidate:R H LiFull Text:PDF
GTID:2438330518492497Subject:Education Technology
Abstract/Summary:PDF Full Text Request
Microblog has become a comprehensive social media platform based on social relations browsing and disseminating of information.The emotional analysis of microblog is a very popular direction in the field of Natural Language Processing(NLP)now,and it is also an important data to pay attention to network public opinion.In the Age of Big Data,digital information resources are growing in an explosive way,how to quickly grasp the general emotional trend,save the cost of analysis,is the purpose of this study.This paper observed that microblog contains a large number of emoticons,the emotional information contained in the expressions should not be ignored,even to a certain extent,represents the feelings of microblog text.If we can divide the emotional category of the emoticons and use the emotional category of the emoticons to represent the emotions of the microblog text in which they are placed,it may save about 16%of the workload of the microblogging emotional analysis.On this basis,this paper presents a method based on emoticons to study the emotional tendencies of microblog,which divides emotions into fine-grained seven categories including happiness,like,anger,sadness,fear,disgust and surprise.Using the Ansj word segmentation tool and the Chinese emotion word ontology of Dalian University of Technology to segment and quantify the microblog text.The emotional value of microblog is calculated and the emotional category and polarity are classified by combining the dictionary and simple rules,then construct emoticon emotional dictionary,and determine the accuracy of emoticon emotional dictionary constructed.The preliminary experimental results show that the accuracy of emotional value and emotional polarity obtained by the combination of dictionary and simple rule is 83.2%after being modified,but the accuracy of the emotion category is too low.Then,a kind of emotion classification based on artificial annotation and emoticons correlation is proposed,label part of the emoticons' emotion category artificially,make rules to classify the emotion category of the part of the emoticons which with high expression of the labelled emoticons,determine the unlabeled emoticons' emotional category iteratively in turn,update emoticon emotional dictionary,and determine whether the whole microblog emotions can be represented by the emotions of emoticons it contains.Ultimately it shows that emotional recognition of types of happiness,like,disgust is the best,the correct rate is more than 90%,correct rate of types of anger,sadness is 80%or more,fear,correct rate of types of fear,surprise is more than 70%,and the experiment finally shows that use the emotions of emoticons on behalf of the emotional category of microblog is feasible.
Keywords/Search Tags:Emoticons, Emotional Dictionary, Fine Grained, Emotional Classification
PDF Full Text Request
Related items