基于演化算法的唐诗自动生成系统研究  

Research on a Tang Poetry automatic generation system based on an evolutionary algorithm

作　　者：穆肇南 刘梦珠 孙界平[2] 王成 MU Zhaonan;LIU Mengzhu;SUN Jieping;WANG Cheng(College of Computer and Information Engineering,Guizhou University of Commerce,Guiyang 550014,China;College of Computer Science,Sichuan University,Chengdu 610207,China;Guangzhou Huizhi Communication Technology Co.,Ltd.,Guangzhou 510630,China)

机构地区：[1]贵州商学院计算机与信息工程学院,贵阳550014 [2]四川大学计算机学院,成都610207 [3]广州汇智通信技术有限公司,广州510630

出　　处：《华东师范大学学报（自然科学版）》2020年第6期129-139,共11页Journal of East China Normal University(Natural Science)

基　　金：贵州省普通高等学校科技拔尖人才支持计划(黔教合KY字[2016]086);贵州省教育厅青年科技人才成长项目(黔教合KY字[2018]274);贵州商学院院级课题(2019YJJGXM08)。

摘　　要：针对基于演化算法的唐诗自动生成系统展开研究.研究工作主要有:使用GloVe(Global Vectors for Word Representation)模型训练词向量,设计了基于关键词和平仄押韵的初始种群方案、基于语法语义加权值的适应度函数、基于锦标赛算法的选择策略、基于启发式交叉算子和启发式变异算子的演化算法;给出了基于演化算法的唐诗自动生成模型及系统实现.实验表明,根据给定关键词,该系统初步实现了唐诗的自动生成,生成的唐诗经人工修改后具有一定的欣赏价值.In this paper, research on an automatic generation algorithm for Tang poetry, one of the poetry genres, is carried out. The research work consists of the GloVe(Global Vectors for Word Representation)model to train-word vectors, an initial population scheme based on keywords and peaceful rhymes, a fitness function for grammatical and semantic weights, and a selection strategy used in tournament algorithms;the latter includes heuristic crossover and heuristic mutation operators as well as automatic generation of Tang poetry based on an evolutionary algorithm. Experiments show that by providing keywords the established model and system can achieve the initial goal for automatic generation of Tang poems. After manual modification, the proposed system can generate valuable and appreciable Tang poems.

关 键 词：演化算法 自动生成 唐诗 自然语言处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]