Software clones are fragments of code or other artifacts that are copied or repeated in source text.  Clones have a significant effect on software development and maintenance, and as a result clone detection, analysis and management has become a popular research topic in software engineering.  However, there is a class of languages for which clone detectors perform very poorly, and in fact return results that are either incomplete, uninteresting or downright incorrect. In this talk I will introduce one such language, the Web Services Description Language.  I will explain why implicit assumptions made by clone detectors cause them to do very poorly in analysis of similarities in this and other modelling languages, and how we can resolve the problem by taking context into account.  I will demonstrate our results by applying the new method to the problem of web services personalization, and show how it generalizes to semantic methods such as LDA.