A Study Of Chinese Poetry Generation Based On Deep Learning

As a literary form unique to the Chinese,the traditional Chinese poetry displays the unique cultural deposits of China with their special formats and rhymes.The traditional Chinese poetry is a vital heritage in the glorious ancient Chinese civilization.With assorted types,the traditional poetry can be divided into Tang Poetry,Song Ci,and others,from the perspective of era;or "ancient poetry" and "modern-style poetry",from the perspective of rhymes.In the natural language processing area,the generation of the ancient poetry is an extremely challenging task.There are two different types of poetry machine generation which are based on the traditional machine learning and deep learning.This paper puts forward a way to generate poetry based on the Chinese Poetry Generation with Planning based Neural Network,introduced a new multidimensional emotion judgement model:first of all,the user confirms the text of the theme for the writing,extracts key theme words from the text,and designs the theme and outline of the poetry based on the key theme words.Then,based on the outline,the coder-decoder model is used to generate the poetry.After that,the poetry generated will be analyzed from the perspective of emotions to judge whether it suits the emotion distribution rules in poetry writing.Those that satisfy the emotional rules are produced while those that do not satisfy will be re-polished to be produced when it meets the requirements.Under the new emotional judgement,the generated poetry is better than the original model and the natural language is more human-like.This paper has the following innovations:1.This paper makes the very first attempt to introduce the multi-dimensional emotion analysis into the poetry generation system and put forward a model to judge the types of the poetry from the perspective of emotions,which can guide the generation model to generate quality poetry that better meet the emotional rules.2.In processing the segmentations in the data processing stage,there are some researchers segments the words based on modern poetry corpus.With the change of the time,the meaning of words has changed a lot.This paper,with the help of the already available ancient poetry corpus,segments the words and refines the granularity of the segmentations,so as to make the meaningful groups of the segmentations loser to the meanings in the ancient poetry.