Semantic Interaction in Data Flow Oriented Enterprise Environments
Visualization is used as a tool to understand the data, to see the unseen. In this process, interaction plays a fundamental role, as it can reveal insights that a single representation or animation does not. One of the most interesting but less studied type of interaction is the semantic one, which consists in interacting with the data by means of the image that represent it. In this case, the user could be interested in the original data that is behind a single pixel or in the set of data that generates an area of the image. The purpose of data identification and selection is not only to get a detailed and specific view of that data subset but also to extract data for remodeling and further calculations [FEL92].
Semantic interaction provides several advantages. First of all, it is more intuitive, as the user interacts with the final image but can think in terms of original values and context [SCH95]. Furthermore, it allows the user to concentrate on the task at hand, instead of focusing in the interface itself, avoiding disruptive switch of the cognitive context. This makes the interface more “transparent”, resulting in the enhancement of the cognitive productivity. Finally, it simplifies the coupling of input and output in the interface, suggested by Ware [WAR04] as a mimesis of the objects of the real world, allowing that a visual object can potentially provide output as a representation of data and also potentially receive input.
However, there are major obstacles to achieve semantic interaction, especially in data-flow oriented systems, as one needs to be able to reverse the transformation of every module of the visualization pipeline. To overcome this difficulty, a variety of approaches are proposed in the literature. One of them is the Visualization Input Pipeline technique described in [FEL92]. Here a backwards pipeline is constructed parallel to the original visualization counterpart. Each backwards-oriented module 'knows' the functionality and the parameters of the original module and inverses its data modification. Some application builders provide a similar feature. AVS and IRIS Explorer, for example, both use an augmented data-flow model supporting image probing facilities based on feedback loops – functions that accept a geometrical position, query the input data and return a value interpolated at the required point; but information of higher semantic level cannot be regained.
The implementation of semantic interaction in data visualization applications used in enterprise environments has and additional problem, given by the type of data used in them, which is usually abstract and high-dimensional, without a pre-determined geometric representation. For this reason, it could be difficult for the user to identify the form of interaction with the features or categories represented in the image. To confront this problem, it is required the design of new visualization metaphors for the creation of visualization's meta-controls susceptible to semantic interaction for data analysis and decision making tasks.
Based in the beforementioned, this paper proposes a new model of (semantic) interaction and new visual metaphors in a new application area of the field of Business Informatics. The model and the interface principles developed will be general enough to be applied in solving a wide range of visualization problems. Based on the obtained model, we will develop a programming environment that facilitates the application of semantic interaction in the type of systems studied.
[FEL92] [FEL92] Felger, W., Schroder, F.: The visualization input pipeline-enabling semantic interaction in scientific visualization. Computer Graphics Forum 11(3) (1992) C139-C151
[SCH95] [SCH95] Schröder, F.: ape - the original datafow visualization environment. SIGGRAPH Comput. Graph. 29(2) (1995) 5-9
[WAR04] [WAR04] Ware, C.: Information Visualization: Perception for Design. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA (2004)