| A Workshop on Machine Learning in Natural Language Processing |
Organizers: Shalom Lapin and Ido Dagan |
Weak Bias Language Models and Universal Grammar
Shalom Lappin
King's College, London
It is widely believed that the scientific enterprise of theoretical linguistics and the engineering of language applications are separate endeavors with little for their techniques and results to contribute to each other at the moment. We explore the possibility that machine learning approaches to natural language processing being developed in engineering-oriented computational linguistics may be able to provide specific scientific insights into the nature of human language. We argue that, in principle, machine learning results could inform debates about the cognitive basis of language acquisition. To the extent that a given machine learning experiment is successful in acquiring understanding of a particular linguistic property, it shows that the learning bias that the model embodies is sufficient for acquisition of that understanding. If, further, the bias is relatively weak, containing few assumptions and little task-specificity, the experiment provides motivation for the view that a weak bias model is adequate to sustain the acquisition of this type of linguistic knowledge. Existing results from both empirical work in grammar induction and theoretical developments in computational learning theory that inform this work offer initial tentative support for a weak bias view of universal grammar. Joint work with Stuart Shieber, Harvard University and Michael Collins, MIT |