Cool Logic

Philip Schulz (ILLC)

A Naïve View on Language

May 3rd at 17:30, in Science Park 107 F1.15

This is a talk about language research. It is not (only) a talk about linguistics. The aim is to give an introduction to modern research in the study of language that should be accessible to everyone. We are going to approach the matter naïvely, meaning that we will try to keep theoretical prejudice to a minimum and rather look at known facts about language. While we will retrieve some of these facts from the experimental literature, but we will also try to exemplify some of them during the talk.

Although language engineers have benefited from a small subset of research in so-called “formal linguistics”, we still know next to nothing about language itself. This is partly because “formal linguistics”, which has been the main-stream in linguistic research for decades, has for a long time been constrained by a set of assumptions that could never be empirically justified. Moreover, researchers working in this field never even considered real data. This has led to a strange opposition: while linguistics claims a status as an empirical science, it has evolved into a battle of opinions amongst theorists.

We will start our alternative investigation by briefly introducing some of the main assumptions that have straightjacketed the investigation of language for too long and give some arguments for why they are wrong or plainly irrelevant to the study of language. Regarding this issue, I will not only present my own opinions, but also frequently cite eminent researchers from both academia and industry. Departing from that point we will look at known facts about language, starting with the simple observation that language is linear in the sense that one sentence/word/syllable follows the other. (Surprisingly, this fact has been consistently ignored by almost all formal linguistic theories.) Based on many experimental and corpus studies – where a corpus is just a large collection of written or spoken language – I will argue that acknowledgement of this simple fact already gives us massive explanatory power. Time permitting, we will review some recent empirical results that extend this line of argumentation beyond linearity and discuss the implications that the approach presented in the talk may have for the study of language and the methods employed therein.