Research On Multidimensional Emotion Classification Algorithm Of Internet Comment Based On Word Vector

Aiming at the sentiment analysis of web text,this thesis proposes a sentiment analysis model based on complex sentences and complex semantics,and applies it to sentence-level multidimensional sentiment classification.First,multi-dimensional sentiment words are expanded on the basis of text corpus in a specific field to build a multi-dimensional sentiment dictionary;Second,the model extracts text feature vectors according to the sentence model and semantic complexity.The text feature vector contains the features of the related words in the sentence model,the features of the emotional subject in complex semantics,and multi-dimensional emotional words.Lastly,Naive Bayes is selected as the classifier for multi-dimensional emotional classification.Experimental results show that,compared with similar models,the model has a good accuracy and recall rate in a multi-class emotional task in a specific field.This thesis presents a model of sentiment discrimination algorithm based on word vectors.The main work is as follows:1.Expanding on multi-dimensional emotional new words in specific fields.The multi-dimensional sentiment words in the basic sentiment dictionary are expanded to construct a multi-dimensional sentiment dictionary in a specific field,so that the sentiment dictionary can reflect the sentiment orientation in a specific field in a deeper level.The same word will have different emotional tendencies in different contexts,so we must filter out the words with emotional tendencies in this field.It is more reasonable to judge the sentiment tendency expressed by the sentence based on such a sentiment dictionary.2.Aiming at the complexity of Chinese sentence patterns,a corresponding sentence pattern model is proposed.Chinese sentence patterns can be divided into several categories according to sentence structure,such as turning sentences,negative sentences,etc.This thesis first extracts the combination of subject words in the sentence,which includes the subject real words and the emotional words before and after;then extracts the feature words of each complex sentence pattern(such as transition words),and divides the sentence patterns into 3 categories according to the transition word categories.By combining sentence subject word combinations and sentence transition words,a complex sentence model is constructed.After the construction of the sentence model is completed,the established multi-dimensional sentiment dictionary and the subsequent multidimensional sentiment thresholds defined by the complex semantic aspects are used to make a comprehensive judgment and classify the text sentiment tendencies.3.In order to verify that the theory introduced in this thesis can be applied in real life,we designed a multi-dimensional sentiment classification prototype system.The prototype system can complete functions such as text crawling,text sentiment discrimination,and emotion category display.