Formal concept analysis in information science

Formal Concept Analysis (FCA) is a method for data analysis, knowledge representation and information management that is widely unknown among information scientists in the USA even though this technology has a significant potential for applications. FCA was invented by Rudolf Wille in the early 80s (Wille, 1982). For the first 10 years, FCA was developed mainly by a small group of researchers and Wille’s students in Germany. Because of the mathematical nature of most of the publications of that time, knowledge of FCA remained restricted to a group of “insiders”. Through funded research projects, FCA was implemented in several larger-scale applications, most notably an implementation of a knowledge exploration system for civil engineering in cooperation with the Ministry for Civil Engineering of North-Rhine Westfalia (cf. Eschenfelder et al. (2000)). But these applications were not publicized widely beyond Germany. During the last 10 years, however, FCA has grown into an international research community with applications in many disciplines, such as linguistics, software engineering, psychology, AI and information retrieval. This shift is due to a variety of factors (cf. Stumme (2002)). A few influential papers stirred interest for FCA in several fields. For example, Freeman and White’s (1993) paper on social network analysis initiated an interest for the use of FCA software among sociologists. In software engineering, several FCA papers (such as Fischer (1998) and Eisenbarth et al. (2001)) won Best Paper Awards at conferences (Snelting, to appear) because FCA happened to facilitate a type of analysis which was previously not available in that field. As Stumme (2002) explains, FCA shifted emphasis to applications in computer science partly due to a merger with the Conceptual Graphs community (Sowa, 1984). An overview of the relationship between Conceptual Graphs and FCA is provided by Mineau et al. (1999). Some of the structures of FCA appear to be fundamental to information representation and were independently discovered by different researchers. For example, Godin et al.’s (1989) use of concept lattices (which they call “Galois lattices”) in information retrieval is based on an independent discovery by