Semantic Driven Visualization

We introduce methods for optimizing the visualization process utilizing semantic information. The semantic metadata are used for data reduction and for emphasizing interesting parts of the data. These methods are also based on determination of Region of Interest (ROI) of the current user. The ROI is derived from the user’s location, his query and the semantic metadata. The visualization process and the process of visual parameters change is described. The system is able to deal with any XML based data formats thanks to the separation of the data model and execution engine. Introduction The classical visualization methods deal mostly with visualization of geometrical features of application data. With growing complexity of application data the urgent needs arise to influence the visualization process by the semantic part of the application data. The use of semantics opens new prospects for development of methods for application data filtering that reduce the complexity of the data and change the presentation aspects to make the data more understandable for the user. Such filtered data can be manipulated more efficiently. Traditional approach of filtering based on geometrical information only [8] is not appropriate for many applications. In many cases it is necessary to handle data from the point of view of their meaning related to particular application the semantics will be used. By means of semantics we can handle also such aspects as ROI (Region of Interest) and DOI (Degree of Interest) in a wider context, not only as a function of geometrical properties of analyzed objects. Semantic based definition of ROI and DOI is used not only for extraction of the needed data, but also for emphasizing of the crucial parts of the data. In this paper we describe a new approach for application data visualization that is based on combination of location awareness, semantic description and the data itself. We introduce new model of application data enriched by semantic description and usage of location information during the visualization process. Use Case In our scenario we work with users who need to get visual information in a mobile environment. The users typically use a notebook or a Tablet PC equipped with wireless connectivity (GPRS). The required data are either 2D map of users' destination in SVG (Scalable Vector Graphics)[2] or a 3D model in VRML (Virtual Reality Modeling Language)[3]. The data are stored in a central database on a remote server accessible through our system. The user’s location or destination is measured using sensors or is explicitly given by the user. The system will search for the required data and will perform data change according to the current needs. The change of data consists at first of data reduction and secondly of modification of graphical representation. Both these operations should help the user to better perceive and understand the data content. The user thus gets the relevant information only. The user’s ROI is moreover visually emphasized to better find the desired information. Visualization process The visualization process in our case consists of three phases: determining of ROI and DOI, choosing graphical representation, performing visualization (see Figure 1). To determine ROI and DOI the context of use (e.g. location of the user) in combination with semantic description of the data (e.g. classification of the information items and relations between them) is used. We divide the data into elementary information items and calculate DOI for each of them. The DOI is represented in a normalized form as a real number between 0 and 1. The DOI values determine the ROI. Semantic ROI, DOI calculation Choosing graphical representation Performing visualization User query Context of use Data (including semantics) Visualized data (including semantic) Figure 1. Visualization process Once the DOI is computed the graphical representation for information items can be chosen. The graphical representation is also restricted by the chosen output data format. In our case we focus mainly on SVG and X3D/VRML. The chosen graphical representation is transformed into visualization commands that are processed during the "Performing visualization" phase. Semantic driven ROI, DOI calculation The complexity of the application data (especially of graphical nature) grows and the efficiency of data manipulation (searching, filtering, editing) becomes a serious problem. One of the possible solutions is to introduce new methods for data manipulation based on semantic description of application data. To be able to define these methods we need a new application data model (ADM) that incorporates semantic description. Our ADM consists of the following information categories: • Geometric information. It contains mathematical description of object geometries. • Visualization information. It consists of a set of graphical information (e.g. color, transparency) and rendering contextual information (e.g. zooming, view transformation). • Structure information. It describes the ordering of objects in the scene i.e. hierarchy. • Semantic information. It reflects the relations between objects that are of non-structural nature (usually related to application). Semantic information can be described in the form of a graph (see Figure 2). The white-filled nodes labeled W1-4 and D1-3 represent real objects that have geometric and visualization information. The gray and black-filled nodes labeled W, R1-2, R, and D represent abstract objects and with their relations (arrowhead lines) describe the semantic information. For the formal description of the semantic information we have chosen the MPEG-7 and OWL formats. The semantic description provides new richer and more precise objects classification. It can be used to specify user’s regions of interest (ROI) based on the semantics. ROI are defined as triples (object1, relation, object2). The element “give-me-all” (represented by “*”) can be used instead of each element in the triple. For example the triple (room1, connection, *) selects all objects that are connected to room1. The semantics can be also used not only for more sophisticated and precise extraction of objects in the ROI, but also for choosing the advanced emphasizing filters that changes the geometrical and visualization information of particular objects based on semantic driven DOI calculation. In such a way filtered and emphasized application data are more understandable for the user and thus increase the efficiency of data manipulation.