基于文本信息的人物性格分析算法的研究与实现

Research and Implementation of Character Analysis Algorithm Based on Text Information

本课题研究基于文本信息分析人物性格,聚焦于静态文本来获取小说、剧本等文学作品中的人物性格,使用《平凡的世界》这一小说著作作为样本分析训练模型,结合心理学中的大五人格算法,主要采用神经网络与传统机器学习相结合的方式,通过对比doc2Vec和word2Vec + CNN两个模型的模型效果发现在预测未知人物性格时前者有着更好的表现,因此,该模型将智能分析文本这一想法变为可能,并且可通过大五人格量表将预测得分映射出人物的性格词汇,使得未来机器能够“读懂”语义,对用户画像、智能机器以及心理学的发展具有重要意义。

Based on the text information, this research focuses on the static text to obtain the characters’ characters in novels, scripts and other literary works. Using the novel “the ordinary world” as the sample analysis training model, combined with the big five personality algorithm in psychology, it mainly adopts the combination of neural network and traditional machine learning. By comparing the model effects of the two models doc2Vec and word2Vec + CNN, it is found that the former has a better performance when predicting the character of an unknown person. Therefore, the model makes the idea of intelligent text analysis possible, and the prediction score can be mapped out the characters’ character vocabulary through the big five personality scale, so that the future machine can “read” the semantics, and the user’s portrait, intelligent machine and psychological development is of great significance.