Abstract: Aiming at solving recognition of Chinese place names, this paper constructs the base of recognition-rule and quantifies the rules to denote the differences of reliability in detecting place names. The leveled strategy is adopted and different recognition methods are utilized by the each level. Evaluation on the open test corpus shows that the recall is 92.23% and the precision is 83.88%.

Key words: Chinese place names recognition; Rule quantification; Automatic segmentation; Chinese information processing

摘要： 以带特征词的中文地名和不带特征词的中文地名作为识别对象，通过构建地名识别规则库，以及对规则库中规则的量化处理来体现规则在识别地名中的可信程度的不同；为提高识别的召回率，采用了两级处理策略，其中每级采用不同的识别方法。开放测试结果表明，召回率为92.23％，精确率为83.88％。

关键词: 地名识别；规则量化；自动分词；中文信息处理