论文部分内容阅读
提出条件随机场(CRF)与规则相结合的地理空间命名实体识别方法。该方法以丰富的知识作为触发条件,用CRF对满足条件的片段作地名及机构名识别,识别出来的命名实体又被解构,CRF及知识用来进一步判断该命名实体是否表示事件发生地的地理空间信息。实验结果表明,统计与规则方法的结合以及解构算法有效提升了地理空间命名实体识别的性能,准确率、召回率和F1值分别达到92.86%、90.91%、91.87%。
This paper proposes a geospatial named entity recognition method based on the combination of conditional random fields (CRFs) and rules. In this method, rich knowledge is used as the triggering condition, and the CRF is used to make a name and place name recognition of the fragment that satisfies the condition. The identified named entity is deconstructed again. The CRF and knowledge are used to further determine whether the named entity indicates the geographical location of the event Spatial information. The experimental results show that the combination of statistical and rule methods and the deconstruction algorithm effectively improve the recognition performance of named entities in geospatial space. The accuracy, recall and F1 are 92.86%, 90.91% and 91.87% respectively.