Mining Opinion Sentences Based On Extended Emotional Dictionary In Micro-blog Environment

Micro-blog has its own text features,such as short content,sparse semantics,information fragmentation and so on.These features bring difficulties to the expansion of emotional dictionary,and more and more new words with emotional tendency emerge in micro-blog,which also makes the mining of opinion sentences more difficult.Aiming at the expansion of emotion dictionary and the accurate expression of topic opinion mining,this thesis proposes the topic opinion mining of micro-blog based on the expansion of emotion dictionary.First,we extend the emotion dictionary based on cooccurrence probability,and then improve TF-IDF algorithm by part of speech weighting and sentence beginning and sentence ending weighting to extract keywords.At last,the thesis extracts,analyzes and summarizes the opinion sentences of micro-blog topic.The main research contents are as follows:(1)Based on the cooccurrence probability,the affective dictionary is extended.Firstly,text data preprocessing is carried out to filter out non text elements in text data.Then through the cooccurrence probability of new words and the emotional probability of new words,we can judge whether the extracted new words can be used to expand the emotional dictionary,and calculate the emotional tendency of new words that have been loaded into the emotional dictionary.Finally,the new words are added to the affective dictionary by iterative calculation until the affective dictionary is no longer extended.(2)Based on TF-IDF improved algorithm and inter word association rule calculation,this thesis mine opinion sentences of hot topics in micro-blog.First of all,the TF-IDF algorithm is improved by part of speech weighting and sentence beginning and sentence ending weighting to obtain the core keywords of the text data;further,the left and right entropy values are calculated to filter the core keywords;finally,based on the extended emotional dictionary obtained in the first step,the emotional values of the core keywords are calculated,and the relationship between the core keywords is mined to achieve Extraction of opinion sentences.The experimental results show that the extension of emotion dictionary and the improved algorithm of opinion mining proposed in this thesis can effectively mine the opinion sentences of micro-blog topics.In this thesis,we extend the traditional emotion dictionary based on cooccurrence probability algorithm,optimize the mining of opinion sentences by association rule algorithm and left and right entropy,accurately mine the topic opinion sentences of micro-blog,and build the representation of the association rule graph of opinion sentences.