Abstract: For non-expert users, the general text association rules are hardly understood, moreover graphical visualization in the traditional sense is just popular for experts in the data mining field. To address these problems, a novel visualization methodology of association rules is proposed based on the Natural Language Generation （NLG）, which introduces NLG technology to the association rules visualization. The proposed approach can interpret the items of rules as ordinary natural language by using the interpretation schema in the domain knowledge base, and ultimately generates the smooth and easy natural language sentences through sentence planning and surface realization. The experiments show that the results obtained by this approach are more easily understood to the non-expert users, and help them make accurate decisions by taking full advantage of value of information gained in the mining process.

Key words: natural language generation, association rules, visualization, domain knowledge base

摘要： 针对传统的关联规则蕴含式表示方式和图形可视化方法对非专家用户来说不易理解的问题，提出了一种新的基于自然语言生成的关联规则可视化方法。该方法将自然语言生成技术引入到关联规则可视化中，通过领域知识库中的解释模式将关联规则中每一项生成简单的自然语言句子，并经过句子规划、表层实现，最终生成流畅的自然语言句子。实验最终得出的结果，便于普通用户理解和应用，从而帮助用户获取更有价值的信息。

关键词: 自然语言生成, 关联规则, 可视化, 领域知识库