面向汉语的自然语言路径描述方法

Approach to Route Description in Chinese Natural Language for Electronic Maps

自然语言路径描述是运用日常生活中人们习惯的语言方式和思维方式来表达路径信息的。而可视化描述和自然语言描述相结合,可以使电子地图向用户提供图文并茂的路径查询服务。目前,电子地图在移动设备和Internet日益普及,自然语言路径描述成为电子地图大众化和人性化服务的重要标志之一。因此,路径描述中需要重点解决空间关系的自然语言描述问题。自然语言中空间关系描述的语言特征与心理学和认知语言学有着非常密切的关系。本文描述了针对汉语的路径描述词汇库的结构与内容,并根据汉语特点设计了路径描述的单元和复合句法模式,以解决词汇受限和句法模式单一的问题,提出了道路名称、沿途标志性地物和方向判断等路径描述补充信息的获取方法,以增加路径描述的可理解性。并以南京市电子地图为例,开发了电子地图查询原型系统。该系统能够在地图上可视化显示最短路径的同时,自动生成该路径的自然语言描述。

Natural language route description represents path information using the ways of linguistic and thinking in people＇ s daily life. Route description in visualization and natural language provides path query service with both pictures and explanations. With the increasing of electronic maps in mobile equipments and Internet, natural lan- guage route description becomes one of the significant services for the popularization and humanization of electronic maps. Natural language spatial relations play an important role in route descriptions, which are related to psychology and cognitive linguistics. This paper aims to solve the problems of generating natural language route description in Chinese way-finding systems, on the basis of datasets of geographical information systems and natural language generation technology. Firstly, based on the introduction of the shortest path algorithm, this paper describes the classification of terms for Chinese route description. Secondly, four unit syntactic models are presented to describe different path units. The combined syntactic model is proposed to describe one complete path by means of linear integration of path units, which focuses on the solution of the problem of limited syntactic models in the previous approaches. Furthermore, the algorithms are discussed for the extraction of additional information such as road names, landmarks and orientations according to a given path, in order to improve the understandability of route descriptions. Finally, one prototype system is developed to implement the proposed approach with the case of naming electronic map. It indicates that our proposed approach can achieve satisfactory performance. This system can simultaneously visualize the shortest path and generate its natural language description based on query requirements