Faculty of Computer Science

Research Group Theoretical Computer Science

Oberseminar: Heterogene formale Methoden

Abstract:

Traditionally, ontologies are developed around concepts and their definitions. This methods has some drawbacks if there is a lot of overlap among concepts.

Formal Concept Analysis (FCA) is a data-driven approach to construct concept hierarchies. Based on a set of objects and their attributes, a concept lattice can be built, using a closed-world assumption. (Lattices can be seen as a weak form of Boolean algebras without negation.) Since the 1980ies, many applications of FCA have been developed in very diverse areas such as ontologies and knowledge representation, information retrieval, frequent item set and association rule learning, data mining, knowledge discovery, text mining, software product lines and others. Recently, word2vec word embeddings into vector spaces obtained by neural networks have been adapted to build fca2vec, a method for displaying large FCA concept lattices in 2D or 3D.