Philosophical foundations of distributional semantics

(Advanced Topics in the Philosophy of Language)
rMA Philosophy/Master of Logic (ILLC) seminar, University of Amsterdam, April-May 2017

Technical developments influence traditional philosophical concepts more than professional philosophers are readily prepared to realize and admit. After computation, it’s (again, in fact) the turn of meaning.

What is meaning? This is one of the big questions in philosophy, in particular in philosophy of language. As it happened in the past, age-old philosophical questions regarding key concepts are all of a sudden revived because of high-octane technical developments in neighboring scientific fields. This has been the case with computation or ontology in computer science, and it is right now the case with the concept of meaning in the new, booming field of distributional semantics.

A strand of semantics originating from structural linguistics, distributional semantics is presently thriving in computational linguistics. Distributional semantics’ key premise is that the meaning of words can be derived from their linguistic context.

But which concept of meaning does distributional semantics exactly presuppose, when millions and millions of words and patterns are crunched by a computer to detect (find out? model?) the meaning of ‘woman’ or ‘black’ or ‘tolerance’? Better put: which concept of meaning should the enterprise of distributional semantics exactly presuppose, to be methodologically sound?

Barring cursory appeals to Wittgenstein’s concept of meaning-as-use in current literature in distributional semantics, a satisfactory reply to these questions is practically absent.

In this course you will engage in serious interdisciplinary reflection on the philosophical foundations of distributional semantics. At the end of the course you will be familiar with the historical background relevant to the theoretical foundations of structural linguistics (Bloomfield, Harris, Firth), and with methodological criteria of theory-choice applicable to linguistics; you will be able to put in both historical and systematic relations little known ideas by Margaret Masterman (1910-1986), a pioneer in computational linguistics; you will master a philosophical toolbox comprising relevant concepts of meaning and their variants, drawing among others upon ideas by Wittgenstein and Quine, and familiarize yourself with technical applications of the ideas just mentioned in current work by distributional semanticists.

Finally, you will demonstrate to be able to present the new knowledge you have acquired clearly and effectively both orally by giving a short presentation, and and in writing, by producing an original, research-level paper.

Readings: Compulsory readings by structural linguists, Masterman (& Wilks on Masterman), Quine, and possibly Wittgenstein; current work in philosophy of linguistics (Stokhof & van Lambalgen; the SEP entry on Philosophy of Linguistics, this paper of mine), and by distributional semantics practitioners (Lenci, Herbelot).


Here you find the study guide for 2016/17.