| With the further development of mobile GIS,intelligent GIS and socialized GIS,the geospatial information service based on natural language processing is an inevitable trend in the field of geographical information science.The intelligent conversion from text to graphics is one of the important research directions.Both natural language and maps have the ability to express geographical entities spatial information.Natural language is easy to use and has a high degree of abstraction,however the language of the map is more intuitive,and it contains rich spatial information.Converting natural language to map language can help people have a more intuitive understanding of the geographic space environment described by the natural language.However,How to make a computer has the same ability as a human brain to construct graphical information from natural language?How to make a computer have the intelligent spatial cognitive thinking?They are the current research difficulties.Taking the geographical entities spatial information "text(natural language)map" intelligent conversion as a goal,this study develops theories and methods for"text-map" conversion concerning shape,size,spatial distribution of geographical entities in natural language description.It can help to build the foundation for the development of theories,methods and applications of a new generation people-oriented GIS.The main research contents and conclusions are provided as follows:(1)Theoratical framework research on geographical entities spatial information"text-map" conversion.The concept,key problems and conversion process of"text-map" conversion are discussed.There are four stages for the "text-map"conversion such as structuralization,quantification,symbolization,and visualization.(2)Quatification research on natural language spatial relations with consideration of the geometrical type of geographic entity.This study proposes a method that using point coordinate pairs,straight line segments and rectangular/circular shapes to quantitatively represent point,polyline and polygon geographical entities in the natural language respectively.In addition,the natural language spatial relations types between point and point,point and line,point and surface,line and line,line and surface,surface and surface geographic entities are summarized.Based on set theory,a method of transforming natural language spatial relationships between different geometrical types of geographic entities into graphical spatial relationships is proposed.(3)Research on symbolization of geographical entity driven by fuzzy morphological description.The classification and expression of geographical entities in natural language are elaborated,and the types of morphological description of geographical entities in natural language are summarized.Based on the visual variables theory,four types of visual variables of geographical entities are extracted from morphological description,which includes shape,size,color,and direction.Then,the representation strategy of geographical entity symbols based on morphological description is studied.In the strategy,semantical ambiguity degree of the morphological description of geographic entities is divided into five levels from high to low,and symbols for all levels are designed and analyzed.(4)Experimental verification and effectiveness evaluation of map visualization.Based on the above research results,a map visualization expression strategy and visual effect evaluation method based on natural language description are proposed.A prototype system is designed to implement "text-map" conversion,and path and scene descriptions are selected as the experimental text to finish the experiment.In addition,a method of the similarity calculation for maps is proposed,and it is used to compare similarities among auto-generated maps,hand-drawn cognitive maps and real navigation maps.The experiment verifies the feasibility and effectiveness of the proposed method. |
PDF Full Text Request