Abstract

I will talk about some converging trends in cognitive neuroscience and computational linguistics, which I believe offer exciting prospects - both for understanding the operation and organisation of our language competence, and perhaps for building more effective language processing applications. On the one hand cognitive neuroscience is becoming more computationally sophisticated in how it analyses higher cognitive processes, while experiments are becoming more ‘ecologically valid’ in that they involve realistic behavioural tasks like story reading or game playing. On the NLP side, early data-driven methods were often “black-boxes” which used “brute force” to solve niche problems. Now computational linguistics is producing more nuanced models, allowing us to inject theoretical insights into their design, and yielding results that are readily interpretable. I will give an overview of these themes presenting experiments that try to find solid empirical basis for language and meaning in language corpora, behavioural norms, and recordings of brain activity.