Music Sentiment Analysis Based On Title And Location

With the rapid development of artificial intelligence and pattern recognition technology and the improvement of people's quality of life,music has become one of the essential communication media in people's lives.The rapid propagation of the Internet has resulted in a large number of excellent musical works.How to classify a wide range of musical songs with different genres and make relatively accurate classifications and use the advanced technology in the computer field to provide users with accurate personalized recommendations has become a hot topic in current research.As a prerequisite for the classification and recommendation of music retrieval,emotion plays an important role in the transmission of musical emotions.However,the emotion of music is often influenced by many factors.It can express the emotion of a song through singing style,music arrangement,lyrics and other angles.For a widely circulated song,the highest vocal degree is certainly not all,but a small part,and strong emotions are often reflected from this small part.Therefore the position of different characteristic words in music lyrics text is also of decisive significance to the expression of music emotions.This dissertation will construct a unique sentiment dictionary from the perspective of lyrics analysis,and apply relevant classification techniques around the important information of position to explore the emotional classification of music.Music emotions are divided into four categories: happy,sad,quiet and passionate.The main research work of this article includes the following parts:By learning the emotional characteristics of the data set,the CHI feature selection method is used to calculate the degree of correlation between the feature words and the categories,and a sentiment dictionary for the data set is constructed according to the ranking results,which avoids the singularity of the emotional dictionary in the traditional training process.Abandoning the practice of sharing a set of emotional dictionaries for all the traditional data sets,this method uses its own emotional dictionary for each data set construction.This guarantees the robustness of the sentiment dictionary to some extent.Experiments compare the influence of emotional dictionaries of different dimensions on the emotional classification performance of music.Taking into account the characteristics of repeated feature words in lyric texts,a lyric sentiment analysis method based on location factors is proposed.The data setcomes from Kugou and Kuwo Music,the method integrates the text title and lyrics text to position the feature words,taking into account the influence of the weights of the feature words appearing in different positions on the lyric-based music sentiment classification,and uses the feature words in different positions of the title,at the beginning,middle,and end of the lyrics text at different positions.Analytic Hierarchy Process(AHP)calculates different position weights.Based on this,it compares the results of multiple classifiers and verifies the effectiveness of this idea in improving the emotional classification performance of music.