Finding Segmentations of Sequences

We describe a collection of approaches to inductive querying systems for data that contain segmental structure. The main focus in this chapter is on work done in Helsinki area in 2004-2008. Segmentation is a general data mining technique for summarizing and analyzing sequential data. We first introduce the basic problem setting and notation. We then briefly present an optimal way to accomplish the segmentation, in the case of no added constraints. The challenge, however, lies in adding constraints that relate the segments to each other and make the end result more interpretable for the human eye, and/or make the computational task simpler. We describe various approaches to segmentation, ranging from efficient algorithms to added constraints and modifications to the problem. We also discuss topics beyond the basic task of segmentation, such as whether an output of a segmentation algorithm is meaningful or not, and touch upon some applications.