Marking And Recognition Of Spatial Information In Natural Language

Natural language is one of the most important tools for people to communicate in everyday life,and it is more suitable for people's cognitive habits.Compared to the structured computer language,unstructured natural language is more easily accepted by people.The barrier-free "communication" between the unstructured natural language and the structured computer language has become the focus and difficulty of the computer field.The natural language contains a large number of geospatial information,such as geospatial objects,the attributes of geospatial objects and the spatial relationships between the geospatial objects.These spatial information is the most direct description and expression of objective existence and its relationship.Compared with the structured geographic information display,the natural language based geographic information service can better meet the people's awareness of spatial information needs.With the continuous improvement of the Chinese text annotation system,extracting unstructured spatial information from natural language and analyzing it by computer system can be well realized.By constructing the annotation structure of geospatial information and mining the syntactic structure of unstructured spatial information in natural language,we can extract the unstructured spatial information in natural language and store it in structured computer language.It is a good way to deduce people's understanding of geospatial information and describe habits,and better realize the "communication" between unstructured natural language and structured computer language;apply it to all aspects of GIS application,improve the quality of geographic information services,and better serve the people's production and life.Based on the existing technical methods of natural language processing and the method of annotating Chinese text,this paper constructs a system based on natural language spatial information annotation system,which is based on the natural language spatial information labeling and extraction,and automatically identifies the spatial relationship in natural language.Language spatial information description syntax rules,conditional random field and stochastic forest model to achieve natural language geospatial information extraction,natural language spatial relationship recognition classification.Around this theme,the main work is done:(1)Based on the "China Toponymic Settlement",the names of the names of the names of the natural language statements are used to provide a basis for the participle of the natural language statements containing the names of the spatial and geographical entities.(2)Analyze the natural language spatial relation description sentence by means of induction and summarization,and construct the syntactic model of natural language spatial relation description;(3)Using the method of statistical method based on stochastic field model to analyze the natural language corpus,(4)Combining the characteristics of geographic information and the characteristics of descriptive language,this paper sets out the geographical information labeling system and labeling specification of Chinese text,labels the geographic information in natural language,constructs annotation corpus,and classifies the vocabulary in the corpus,The construction of the geographical entity name dictionary,spatial topology relations dictionary,spatial relationship dictionary,spatial distance relationship dictionary;the word frequency for statistical analysis,people get the general rules of spatial information cognitive expression;(5)Randomly extract the statements in the corpus as the training data,summarize the feature sets of the spatial relations types,construct a random forest model for spatial relationship information identification classification,and test them by randomly selected test statements.