Research On Algorithm Of Emotion Semantic Matching Between Music And Text

Natural language and music are two semantic symbol systems for people to express emotions and describe things.The analysis and establishment of semantic association between language and music can not only help to provide more accurate text and music retrieval and recommendation services,but also help researchers to further understand the emotional semantics.Previous studies mainly focus on the surface symbolic features of natural language and music,and seldom consider their semantic meaning,which limits the accuracy and interpretability of some applications based on the semantic association of natural language and music.On the other hand,some applications,such as multi label classification of music emotion,automatic annotation of music,need more accurate semantic association between natural language and music.Therefore,the analysis and establishment of semantic association between natural language and music can promote the application of text and music.Semantic matching of symbols is a key problem in natural language understanding.A large number of natural language processing tasks,such as machine translation,automatic question answering,information retrieval,can abstract idiom meaning matching to a certain extent.The symbolic semantic matching between music and text is different from that between languages.The components and semantic scope of the symbolic system of music and text are different.Music is composed of rhythm,melody,strength and other elements,while text is composed of words,words,sentences and other elements.Emotion is the main semantic of music expression,and the semantic range of text expression is wider,including emotion.According to the semantic features of natural language and music,this paper puts forward the calculation method of emotional semantic correlation degree between natural language and music.The main work of this paper includes the following three aspects:(1)Constructing the model of emotional semantic matching of music clips and words: DT-GRU(Deep Transition-GRU).The model is based on the encoder-decoder structure,and realizes the emotional semantic matching of music clips and words through GRU algorithm and attention mechanism.Through multiple experiments and comparison with other similar models,it is proved that the model can achieve more reasonable emotional semantic matching of music clips and words.(2)Constructing the emotional semantic model of music and text:MT-Attention(Music Text-Attention).The model uses Bi-GRU algorithm,attention mechanism between models and emotion classification to complete emotion semantic matching between music and text.(3)Text music cross modal retrieval.On the basis of(1)and(2),the cross modal retrieval from text to music is visualized.