Abstract

In the last decades, experimental philosophers have introduced the notion that conceptual analysis could use empirical evidence to back some of its claims. This opens up the possibility for the development of a corpus-based conceptual analysis. However, progress in this direction is contingent on the development of a proper account of concepts and corpus-based conceptual analysis itself that can be leveraged on textual data. In this essay, I address this problem through the question of similarity: how do we evaluate similarity between two concepts, as similarity relates to identity? After a survey of prominent conceptual analysis methods, I propose a cursory account of corpus-based conceptual analysis. Then I formulate the question of similarity, and argue for an account that is functionalist in Millikan's (1984) sense. In this process, I propose a new account of concept that bases itself on millikanian teleosemantics in order to account for concepts' contribution in discourse. I then illustrate its fruitfulness by showing how it enables accounts of concept presence detection in textual data, both automatically and by a human judge.