Bilstm Personalised-Style Poetry Generation Algorithm
Zhuofan Sun, Zhengyuan Zhang, Mingxuan Zhang · 2023
Chinese poetry plays an important role in people's daily lives, promoting an understanding of history and culture, enriching their emotions, and enhancing their aesthetic skills. It is an integral part of China's cultural heritage. In natural language processing, Chinese poetry generation is challenging because its unique rhyme format requires expressing certain meanings and achieving perfect harmony between syllables, tones, and rhymes. An improved long and short-term memory network framework is used in this paper to generate each line of the poem. The innovation of this paper is that a style dictionary is provided for different dataset sources, the generated poems are scored based on the style dictionary for their similarity to different styles, and the final poem with the highest similarity is then generated. The results of our comprehensive evaluation of human judgments indicate that our proposed method improves the similarity of the generated poems to the target style and is somehow comparable to the quality of human poetry.