Research On Chinese Novel Emotion Based On Dictionary And Machine Learning

Books are the ladder of human progress.As a spiritual food,books have a huge impact on people.In this ever-changing technology era,how to choose excellent works from a large number of books is important for readers.How to create outstanding works stands out from the crowd is also worth thinking about.Exploring the common model of successful novels and objectively understanding the model is an effective way to solve the above problems.As a common literary genre,the novel has a very high research value.Based on the novel,the dynamic change of emotions is the theoretical basis of the development of the plot,From the perspective of computer,this paper explores the dynamic emotional curve of the novel as the starting point combined with the random fractal theory.The traditional text sentiment analysis mainly focuses on the classification and labeling of emotional polarity of short texts.The emotional research on long texts is rare.The corpus suitable for emotional analysis of Chinese novels is lacking,and the traditional sentiment analysis stays in the results.Induction and summary,lack of further exploration of the research results.In view of the above deficiencies,the work of this paper is as follows.(1)For the lack of research corpus,combined with the current situation of sentiment analysis research,its own research conditions and novel expression techniques,and close context.Based on the"Emotional Vocabulary Ontology Library"published by Dalian University of Technology,this paper expands and adjusts from the following three aspects:1.Using Word2vec to construct word vectors,calculate the cosine similarity,semantic similarity and PMI of word vectors.Achieve new word discovery.2.Constructing the emotion-image library of the emotional research of exclusive novels.3.Use existing emotional resources to match words to expand vocabulary.In this way,the Chinese novel sentiment dictionary is constructed.(2)This paper introduces the random fractal theory to deeply interpret the emotional curve of the novel.The adaptive fractal analysis method is used to remove the overall trend of the emotional curve,and the power-law relationship between the wave scale and the residual is investigated.The Hurst exponent is calculated to characterize the long-range correlation of the curve and use it as a feature to explore the common mode.In order to prove that the conclusion has certain rationality,this paper analyzes the Hurst parameter and the Douban score of the book and the sales volume of Dangdang.The experimental results show that the improved emotional dictionary method can effectively expand the vocabulary and make it more accurate to capture the emotional changes.The eigenvalues of the emotional dynamic curves of 94%excellent Chinese novels are greater than 0.5,which indicates the success of the novel.Emotional dynamics generally have a common pattern of sustained long-range correlation,while providing a mechanism for explaining the reasons for the success of the novel from a dynamic perspective.The experiment verifies that the Hurst parameter and the watercress score have a strong positive correlation with the book sales.This proves that the Hurst parameter can be used as a reference index for objectively measuring Chinese excellent novels and has certain rationality.