结合标注的中文地址匹配规则链模型

Chinese Address Matching Rule Chain Model Combined with Annotation

摘要: 现有的中文地址匹配研究方法集中于对文本特征的研究，忽略了中文地址所包含的建筑特征、地理位置特征、统计特征和行业特征的数据，此类特征数据可以有效辅助中文地址的匹配研究。本文主要面向非规范的中文地址，以燃气行业居民用户数据为样本数据进行实验，通过分析两个数据源中用户信息的多个特征数据，提出以结合标注的中文地址匹配规则链模型。规则链的优点是链内的规则可以动态配置，通过人工和计算机结合的方式，动态管理规则，多次迭代，逐步提升匹配率。实验结果表明该模型可以一定程度提高中文地址匹配的成功率。

Abstract: Existing Chinese address matching research methods focus on text features, ignoring the data of architectural, geographic, statistical and industrial characteristics contained in Chinese addresses, which can effectively assist Chinese address matching research. This paper mainly aims at non-standard Chinese addresses, and takes gas industry resident user data as sample data to experiment. By analyzing multiple feature data of user information in two data sources, a Chinese address matching rule chain model is proposed to combine labeling. The advantage of rule chains is that the rules in the chain can be configured dynamically. By combining manual and computer methods, rules can be managed dynamically and iterated several times to gradually increase the matching rate. The experimental results show that the model can improve the success rate of Chinese address matching to a certain extent.