The Research Of Genetic Algorithms-based Computer Automatic Chinese Ancient Poetry Generation

The computer simulation of the human poetry–a special type of literature and a typical phenomenon of human creativity– is a great challenge of natural language generation. This paper aims to conduct an elementary research in Chinese ancient poetry generation. We analyze the possibility of the computer automatic poetry generation and discuss the method of its implement in detail. Our research includes the following aspects:1. Make a summarization of the machine poetry generation research. We classify the methods and introduce the representative systems of each sort. Make a summarization of computational linguistics-based Chinese ancient poetry research.2. Build the Song Poems corpus and database. According to the rules and forms of Song Poems, sentences are divided into sub-pieces. Closely combined two-character words are extracted by calculating the frequency and collocation rate. After completing the segmentation, we get a lexicon. Strategies such as conditional probability are used to implement automatic pinyin - tagging.3. Establish the grammar criterion of Chinese ancient poetry generation. Propose a Deterministic Finite Automata–based method to judge the grammatical validity. Combine Latent Semantic Analysis and Mutual Information methods to calculate lexical relevancy; Use corpus statistics and Keenage to do the lexical similarity computation; And retain the expert panel to do the stylistic and emotional measurement of words.4. Propose the Genetic Algorithms approach to Chinese ancient poetry generation. According to the characteristics of Chinese ancient poetry, we design the Level and Oblique Tones-based coding method, the grammatical and semantic weighted function of Fitness, the Elitism and Roulette combined selection operator, the partially mapped crossover operator and the heuristic mutation operator. 5. Construct the Genetic Algorithms-based Chinese ancient poetry generation system. Describe its implement, give the flowchart and some instances of the result, and analyze the result.As is shown by a certain number of tests, the system constructed on the basis of the computing model designed in this paper is basically capable of generating Chinese ancient poetry, and we hope that this work can serve as the foundation for further research in the field.