STUDY ON CONVERSION FROM SPATIAL INFORMATION TO NATURAL LANGUAGE
F. Ren, Q. Du, X. Wu
Geographic Information Science, Wuhan
Map and linguistics are two traditional disciplines. And spatial information and natural language are two symbol systems which describe the objective world. Meanwhile, spatial information and natural language is equivalence of human spatial experience respectively in graphic and literal symbol world. With the rapid development of cognition theory and information technology, spatial information and natural language closely related. For a long time, map is regarded as one of important media for information interchange between spatial information and natural language symbol system. In this paper, the author argues that spatial conceptions in linguistics and linguistics model in spatial information lay a sound foundation for the conversion from spatial information to natural language. Furthermore, Geographic Information System and Natural Language Generating provide such conversion with corresponding methodology bases. Consequently, the conversion is not only feasible, but it can break a new path for spatial information service.
In the paper, based on the metaphor of spatial information and natural language that is one of the basic human communication tools. Theoretically, the author systematically generalizes relative theoretic research achievements of spatial system and spatial information modeling in natural language, and comprehensively discusses the theoretical bases of conversion from spatial information to natural language. In methodological view, the author gives out rounded analysis on spatial object model and its spatial relation, basic method system of natural language generating. The main research work is divided into parts as follows:
Firstly, from the macro perspective, the author analyzes the relations between spatial information and natural language which have common basis of spatial cognition and computability.
Secondly, an overview of research achievements in spatial azimuth of modern Chinese by linguists is made.
Thirdly, from semantic view, the paper introduces middle semantic model and adds a transition layer as semantic basis between spatial object and natural language, and then solves the contradiction among implementation approaches and concrete contents of spatial information, links their features at deep level by the middle semantic model, which is formed through semantic abstract and extension of object. Spatial Point of Interest is just an application object of the middle semantic model.
Fourthly, based on syntax, the paper establishes overall framework, which is syntax basis for the conversion from spatial information to natural language and is also a mapping of their surface features based on the middle semantic model.
Finally, from a technical achieving perspective, spatial scene and route description is explored based on technical approach of building knowledge base, computing rules, data dictionary, and text generation.