Final Report Summary - INTERACTIVEX (Human-computer interaction models for interactive exploration of behaviour datasets) Today, there is a rapid growth in the availability of large data sets about the behaviour of dynamic systems (e.g. networks of sensors, telecommunications, the internet, etc.). Examples include networks of geographically distributed sensors in river basins or systems that models traffic behaviour using the global positioning system (GPS) in smart phones. Analysis of these types of data is useful for many tasks, including trend analysis, planning, and problem identification. The objective of 'Human-computer interaction models for interactive exploration of behaviour datasets' (INTERACTIVEX) project is to develop models of human-computer interaction for exploratory data analysis. The research is focused on the information presentation part of the interaction, especially in the following innovative issues: (1) new tools for producing natural language presentations for sequences of events, measures, and time; (2) new methods for data abstraction and summarisation over large data sets; (3) a solution for automatic coordination of text and other mixed media output (e.g. charts, timelines, maps, animations, etc.) for large-scale presentations; (4) the approach emphasises reusability of solutions with general components and methods that can facilitate the development of new systems. The project receives funding from the European Union (EU)'s Seventh Framework Programme (FP7/2007-2013) under Grant Agreement PIOF-GA-2009-253331 (Marie Curie International Outgoing Fellowship for individual European researchers). In the INTERACTIVEX project, the individual European researcher is prof. Martin Molina from Spain. The duration of the project is 24 months, including an initial outgoing phase (12 months) in the United States of America (USA) at the host organisation AT&T Labs and a return phase (12 months) in Spain at the host organisation Technical University of Madrid (UPM). As a result of the project, a general solution has been proposed to help users to explore behaviour datasets. This solution is conceived as a web-based digital newspaper with automatically generated news stories that with a presentation style that can help non-expert users to analyse datasets. We have designed general software architecture to support this type of user interface with the following main components: a discourse planner, a data analyser, a natural language generator and a graphic generator. The architecture has been validated in different domains including moving sensors (emergency vehicles, ships, etc.) and static sensors (e.g. hydrologic and meteorological domains). The proposed solution can facilitate developers to construct web applications that help general users to understand the meaning of large volume of behaviour datasets as they are recorded by sensor networks. Online solutions that automatically construct textual or multimedia explanations of the meaning of sensor data using non-technical language can facilitate access by non-expert users with any level of knowledge and, therefore, can increase the utility of sensor network infrastructures.
