摘要
人们在户外行走时,通常使用自然语言表达其对自身位置的感知,在计算机中如何将定性的自然语言位置描述转换为定量的图形语言是地理信息系统、人工智能等领域面临的挑战。该文提出一种基于行人位置描述的道路模拟表达方法,将自然语言位置描述转换为道路模拟地图。首先设计一组空间认知实验,探讨道路上的行人在户外如何使用自然语言描述当前位置;然后基于行人的每条自然语言位置描述估算该位置的候选区域,在候选区域中随机产生该位置的一对坐标值;最后将行人在路上经过的多个位置点坐标数据组合,采用直线拟合和概率拟合两种方法对该道路进行模拟表达。实验结果表明,该文方法能达到较好的模拟表达效果,为自然语言到图形语言的转换研究奠定了基础。
People usually use natural language to describe their perceptions of location when walking outside.Geographic information system and artificial intelligence are facing the problem that how to convert the location descriptions with qualitative natural language into descriptions with quantitative graphical language by computer.This paper presents a road simulation representation method based on location description of pedestrians,which transforms location descriptions with natural language into a road simulation map.Firstly,a set of space cognitive experiments were designed to discuss how pedestrians use natural language to describe the current position in outdoor environment.Then,on the basis of each location description from pedestrians,the candidate region was estimated.Meanwhile,apair of coordinate values corresponding to the location was generated randomly in the candidate region.Finally,with the combination of multiple position coordinates,the road was simulated by two methods,namely straight line fitting method and probability fitting method.The results showed that the method mentioned above was feasible.The position estimation method could effectively estimate the geographical coordinates corresponding to the natural language position description with different fuzziness levels.The road fitting method could effectively reflect the characteristics of linear road such as trend and shape.This paper lays a foundation for the conversion study of natural language to graphics language,which also makes an effort for the development of intelligent human computer interaction based on GIS and the new generation of people-oriented intelligent GIS.
出处
《地理与地理信息科学》
CSCD
北大核心
2017年第6期8-13,共6页
Geography and Geo-Information Science
基金
国家自然科学基金项目"基于自然语言交互的地理空间信息认知与模拟表达研究"(41571382)
国家自然科学基金项目"向量组合学习框架下基于依存混合树的中文语义解析研究"(61472191)
江苏省高校自然科学研究重大项目"基于自然语言交互的地图模型近似转换机制与可视化研究"(15KJA420001)
关键词
自然语言
空间认知
位置估算
道路拟合
natural lang uag e
spatial co gnition
position est imatio n
ro ad fit ting
