Domain-oriented Emotional Semantic Pattern Recognition And Its Application

Since the 21 st century,the emotion analysis,a branch of natural language processing(NLP)has become a hot spot of research.Emotion analysis plays a vital role in online public opinion monitoring,commodity evaluation feedback and political election prediction.Based on the hypothesis that Chinese emotional text analysis contains personality(domain emotional lexicon and emotional model is the typical content of the domain of personality),this paper establishes the research framework of Chinese text emotion analysis according to domain emotion lexicon,domain emotion pattern file and domain emotion self-learning algorithm.The main innovation work of this paper includes:(1)This paper presents and implements a closed-loop self-learning algorithm based on emotion patterns.The algorithm consists of five sub-modules: one core module(dynamic pattern matching module),four auxiliary modules(emotion pattern extraction module,sentence emotion tendency judgment module,emotion word discovery and selection rule module,emotion word filtering module).The algorithm can iterate with field corpus being input constantly,update the emotion pattern set,and identify and extract the emotion word.At the same time,the algorithm has memory,and is able to trace the intermediate process and the results.(2)This paper adopts the modular design method to divide the whole self-learning system into five parts,namely reptile module,corpus cleaning module,database storage module,file system storage module and core algorithm module.The experiment platform based on emotion-mode-self-learning is realized.And the underlying data structure of the database storage module is determined by the structure of the original emotion lexicon.Furthermore,the emotion lexicon based on emotion system is established creatively.(3)The four experiments in this paper come to the following findings: First,the domanial emotional words(named entities,positive emotions and negative emotions)and emotional patterns possess personality in some fields.Second,the general emotional mode applicable for all fields does not exist.The self-learning algorithm based on domanial emotional patterns does not converge in universal case.Third,the algorithm system can effectively identify emotional patterns and emotional words in Chinese texts.Fourth,when the emotional part of speech tagging is used to identify the domanial emotional words,it should be classified in terms of fields.Also,the self-learning algorithm based on domanial emotional patterns should be used to improve the Recall Rate,Precision Rate and F-Measure.