Research On Theories And Methods Of Spatial-temporal Narrative Visualization

With pervasive tendency and popularization of map,the map has played a more and more import role in our life.Meanwhile,the customer-oriented map services is replacing the expertoriented and domain-specific professional products to offer location-based service.However,there are still lots of problems to be resolved in this progress.For example,how to change the research interest from the geographic objects in the “space” to the variety events taking place in the “time”,how to explore the complex,fuzzy and diversity of events in the spatial-temporal reference frame,how to emphasize the important role of human and establish visualization methods according to human's common sense cognition.All of the above are key problems to adapt to the tendency of the popularization of Cartography.In this background,the research on spatial-temporal narrative visualization has been put forward.It takes different narrative forms such as history novels,biography and news text as data sources,and aims to build up a complete set of theory,methods and technology for events information retrieving and visualization analyzing by combining narrative and geospatial technology,so as to promote to merge Geography,Cartography into humanistic social science fields.For this purpose,this paper is focused on improving the events information mining and standardization process,establishing mechanism for mapping textual space into geographical space,and building up scenario-oriented multi-dimensional visualization methods.As a result,the theoretical and technical framework has been established,and the multidimensional visualization methods system including the spatial-temporal dimension,the events characters dimension,emotion dimension and events theme dimension have been built up.Finally,two applications have also been realized to validate the feasibility of all these methods.The main works in this paper are as follows:1.Summarizing the current research status on narrative in Geography and Cartography.The related research achievements such as geo-narrative,geo-spatial narrative,story map and narrative atlas has been analyzed.Based on this,we found that the concept and research category is considered to be unclear,lots of problems such as the theoretical frame,events storage model,mechanism for mapping textual space into geographical space,and the visualization methods system and cognition research need to be further explored.Finally,the research purpose and significance on spatial-temporal narrative visualization has been proposed.2.The theoretical and technical framework has been put forward.First,the research category of spatial-temporal narrative visualization is defined and a narrative-oriented tetrahedron model is proposed to describe the features of narrative object.Then,two events classification systems are established including narrative text-oriented system and scenario-oriented system.Finally,by combing techniques of natural language processing,machine learning,Cartography,cognitive science,information visualization,and knowledge visualization,using both qualitative and quantitative analysis methods,the technical route which includes narrative text mining,textual space mapping into geographical space,visualization methods and cognitive research have been carried out.3.The mechanism for mapping textual space into geographical space has been proposed.Event-based multistage storage model(EMSM)and event elements co-occurrence model(EECM)have been designed based on integrating narrative spatial-temporal semantics.EMSM can divide the events into further details at multi-scale and describe the time-topology relationship,logical structure relationship and casual relationship between events,which can satisfy the needs of multiscale visualization.EECM is designed for exploring and visualizing the relations of events elements(events characters,locations,and details).Meanwhile,based on geocoding and gazetteer,a location match model with human assistance is put forward for encoding the locations described in text to geographic coordinates.4.Multi-dimensional methods for spatial-temporal narrative visualization are categorized and scenario-oriented visualization method system is created.According to narrative information tetrahedron model,dimensions of spatial-temporal narrative has been divided.Different visualization methods have been studied respect to spatial-temporal dimension,events characters dimension and events themes dimension.The applicability of visualization method to different scenarios is analyzed with examples.Several key problems in the application of visualization method in spatial-temporal dimension are also discussed,which is the core of narrative visualization.5.Research on cognition of spatial-temporal dimension visualization methods has been carried out.Based on the questionnaire survey,sketch map,eye-tracking and think-aloud protocols experiments,several key problems in spatial-temporal dimension visualization such as timecartogram interpretation,the usability of space time cube in multi-scale and qualitative location cognition have been studied.Based on this,the corresponding visualization strategies have been proposed,which can be used as guiding principles.6.Visualization of historical events and online news events have been implemented by taking the Long March event and MH370 event as examples,respectively.The former is a long time span historical event with multiple sub-events written in literature style.The framework of historical toponymy mapping,toponymy disambiguation,and time information standardization has been proposed.Series Long March events information have been stored in EMSM model.The latter is a short time span event based on formal narrative style from large text corpus and many characters with complicated relations are involved in this event type.According to these features,a series of methods including events elements retrieving and standardizing methods,the placeoriented events themes keyword extracting and visualization methods,the emotional intensity computing and visualization methods with considering text location attribute have been proposed.Based on EECM model,the feasible technical solution for exploring relationships between event places or characters so as to find the underlying patterns and unknown factors in events has been carried out.