USER INTERACTION

To implement the EFAR information/knowledge discovery approach, two key elements are required in the interaction between the user and the data analysis platform: 1) the ability to apply domain knowledge and expertise to guide the analysis process, and 2) the ability to rapidly understand the meaning of the information extracted from data by translating and visualising it according to the semantics of the problem domain.

The strength of EFAR partly resides in its knowledge-guided exploration and analysis of the data space. Contextual information and experience-based knowlegde can provide criteria which significantly narrow down the analysis of large volumes of data, streamline it in well-defined directions and yield outcomes of high utility.

To steer the analysis process, experts need to be able to easily express facts and conditions which are part of their hypotheses in familiar terms from their domain of expertise. Electrical engineers think about 'spikes in signals', marketers talk about 'hikes' in sales, economists talk about 'boosts' in domestic output. The EFAR platform aims to achieve this through a so-called high-level Hypothesis Specification Interface (HSI), which allows domain experts to express hypothesis components in terms familiar to their domain of expertise. The high-level description is then translated into specifications which drive the data exploration/analysis components.

The reverse capability is equally important: visualising the outcomes of the data exploration and analysis in terms familiar to the domain expert. Whenexpertss look at data, they actively search for specific feature type they are familiar withs – the 'alphabet' which allows them to 'read' phenomena and behaviours. The more difficult or the less convenient it is to read the data from a semantic perspective, the less complex the hypotheses which the analyst is able to follow and analyse. Within the EFAR framework we are developing an approach which allows data to be semantically compiled in a format which makes it easy to comprehend by the domain expert, which allows him/her to deal with an increased number of observations and to rapidly update his/her working hypotheses.