Integrating user preferences in the automatic quality assessment of hydrographic surveys
Résumé
The goal of hydrographic surveys is to measure and describe features which affect maritime navigation, marine construction, dredging, and other human activities at sea. These surveys serve among other thing to create nautical charts. The International Hydrographic Organization (IHO) published recently an update of the Standards for Hydrographic Surveys. It shows how important data quality is for these surveys, by setting new standards which allow to refine the data quality levels of criteria according to stated hydrographic needs. This means that depending on the final use of a survey, its quality could be perceived differently by users. It follows from this that, evaluating the quality of a survey, while taking into account the needs of a the end user, as well as automating this evaluation, are real issues. Next to that, data and information quality are classically examined through multiple dimensions. The choice of these dimensions, as well as their relative importance in the overall assessment of the quality, do obviously depend again on the end user. These dimensions might also be influenced by the context of the survey, which in turn might be perceived differently by each user, and therefore modeled in various ways. Given the many elements that impact the quality of hydrographic surveys, we consider the following research question: How to integrate the end user's perception of the quality of a survey with the quality of the acquisition conditions, while guaranteeing a traceable and explainable overall assessment of the survey ? To answer this question, we propose a process which first uses data quality analysis techniques to automatically determine, from data and metadata, various quality measures of hydrographic surveys, and combine these with an MR-Sort preference model to integrate various user profiles into the overall assessment. We show that these multiple quality dimensions (or criteria), combined with different user profiles (acoustician, cartographer, etc.), can lead to different overall evaluations of the quality of hydrographical surveys. Last but not least, we show how these combined tools can lead to explainable outputs, which can be used to analyse existing surveys, and make recommendations to improve future ones.