Analyzing manuscript traditions using constraint-based data mining

Data mining tasks and algorithms are often categorized as belonging to one of a few specific types: clustering, association rule discovery, probabilistic modeling, etc. For some time now, it has been recognized that concrete tasks do not always fit nicely in this categorization. The concepts of constraint-based data mining and inductive querying have been proposed to alleviate this problem; they offer more flexibility with respect to specifying the task. In this paper, we illustrate an approach that goes one step further: we show how a general-purpose declarative modeling language can be used to specify and solve data mining tasks in the area of philology. These tasks have the following properties: they are easily described in words; they are of real interest to philologists; they cannot be performed using standard querying or data mining systems; manually programming a solution for them is challenging, time-consuming and error-prone. We show that a prototype declarative programming framework, IDP, allows for easy modeling and efficient solving of these tasks. We conclude from this case study that the declarative modeling approach to data mining has a large potential and deserves further investigation.