ISEE : a system for event explanation in hybrid environments - implementation on sensor data associated with document corpora
Nabila Guennouni · HAL (Le Centre pour la Communication Scientifique Directe) · 2022
Due to their great potential in the improvement of safety, comfort, productivity and energy-saving, smart connected devices have become ubiquitous in our everyday lives, impacting different domains from hospitals, malls, farms and buildings to cities, vehicles and electrical grids.In order to boost even more life quality in smart connected environments, many applications offering services based on sensor data have emerged. Event detection is one of these services (e.g. fire detection, stroke detection for patients, air pollution detection). Generally, when an event is triggered in a connected environment, the natural reaction of a manager is to try to understand what happened and why the event was triggered. To find answers to these questions, the traditional approach consists in manually querying the different data sources (the sensor network information system and the document corpus information system), which can be very tedious, time consuming and requires a huge compilation effort.This thesis aims at explaining events detected in connected environments, and more precisely those that occur in environments with heterogeneous information systems (IS) (document IS and sensor network IS).We propose a framework called ISEE (Information System for Event Explanation). ISEE is based on:(i) a model for event definition in heterogeneous environments, which allows users to define the events they want to detect according to different axes of description (document and sensor network); (ii) a process for the interconnection of the semanticnetwork, its role is to exploit the data from the events (definition and triggering) to build context-aware (event explanation) interconnections. These interconnections will be used to bring together the different sources of information and guide the explanation search process; (iii) a model inspired by the 5W1H (What, Who, When, Where, How, Why) approach to structure the explanations in a simple, intuitive and easy to understand way.We propose a generic solution that can be applied in differentdomains. Nevertheless, three experiments have been conducted to validate this proposal in the context of a smart connected research building.