Research And Implementation Of Theme Emotion Analysis Based On Ancient Poetry

Ancient poetry,as a treasure of Chinese culture,not only reflects the spiritual life of contemporary people in the cultural construction,but also plays an important role in the cultivation of modern people’s sentiment and cultural accomplishment.Therefore,the research value of ancient poetry can not be underestimated,but also has long-term research significance.With the rapid development of the information age,more and more scholars begin to study the related work in the field of natural language processing,and the data processing also uses modern computer intelligent technology to replace the traditional manual work.As one of the classic problems in natural language processing,text classification technology can effectively classify massive text data automatically.Although it has been widely studied and applied in real life,it is rarely applied to ancient poems.This dissertation mainly studies the sentiment classification and theme analysis of ancient poetry,and the main works and innovation points are as follows:(1)For the same sentence in the expression semantics,ancient and modern text are obviously different,segmentation as the first step of text processing is also a key step,so to ensure the effective division of words and word meaning vectors in ancient poetry,this paper proposes to use a language toolkit,It is a NLP(natural language processing)toolkit specifically for ancient Chinese processing;(2)The data set of sentiment classification in this dissertation adopts the public evaluation database of sentiment classification of Chinese poems provided by artificial intelligence of Tsinghua University.According to the characteristics of ancient poems,the existing resources of sentiment words are summarized and sorted to judge the sentiment tendency of ancient poems,and the accuracy of the model reaches 78%;(3)The data set of topic analysis is to collect the data of poem induction from poetry websites by using crawler technology,and to mark the data set by category.In this dissertation,we studied a variety of deep learning models,including Text Convolutional Neural Networks(Text CNN),LSTM(Long Short-Term Memory)and Attention mechanism,and used a variety of network structure models with different combinations.It is expected to obtain better classification results,and the experimental results show that the accuracy of the model combining text CNN,Bi LSTM and Attention mechanism is up to 71.24%,which can improve the effect of topic classification to a certain extent;(4)In this paper,the poems of emotional problem and topic analysis problem for classification problem is studied,and based on the above research,this paper established a based on the theme of the ancient Chinese poems emotional analysis system,this system mainly includes the theme analysis module,emotion classification module,query module,history module and user management module five modules.