Model and Domain Independence : an Experience in Model Management and Information Extraction

Abstract

Information systems play a crucial role in handling the ow of informa- tion generated, exchanged, acquired and stored nowadays. The strife to pursue model- and domain-independent approaches by prescinding from the speci city of their respective instances is a natural way to keep advancing towards more scalable and interoperable information systems, and therefore contributing to the creation of a more cohesive and productive world. In this regard, this dis- sertation addresses the problem of model and domain independence applied to two critical elements of the whole lifecycle of information management: the modeling and design phase, and the acquisition and storage phase, respectively exempli ed here by the areas of Model Management and Information Extrac- tion. Speci cally, in this Thesis we rst focus on the topic of model-independent schema and data translations, where we show an extension of a model manage- ment operator at both the conceptual and the operative level, by introducing the concepts of inheritance and polymorphism within its underlying data dic- tionary and the rules used to perform the actual translations. Subsequently, we shift our attention towards the automatic discovery of abbreviations from full- text scienti c papers, where we propose a domain-independent methodology to identify and resolve acronyms and abbreviations and match them with entities of a given domain, all within a suitable implementing framework. Eventually, we discuss potential future directions of this research, by introducing the con- cept of semantic similarity among ontologies and the challenge of automatically build them and align them, in order to both extend the proposed methodology and tackle wider areas of knowledge discovery.