1. The Tradition of Categoricity and Prospects for Stochasticity
2. The joys and perils of corpus linguistics
3. Probabilistic syntactic models
4. Continuous categories
5. Explaining more: probabilistic models of syntactic usage
6. Conclusion:
There are many phenomena in syntax that cry out for non-categorical and probabilistic modeling and explanation. The opportunity to leave behind ill-fitting categorical assumptions, and to better model probabilities of use in syntax is exciting. The existence of ‘soft’ constraints within the variable output of an individual speaker, of exactly the same kind as the typological syntactic constraints found across languages, makes exploration of probabilistic grammar models compelling. We saw that one is not limited to simple surface representations: I have tried to outline how probabilistic models can be applied on top of one’s favorite sophisticated linguistic representations. The frequency evidence needed for parameter estimation in probabilistic models requires a lot more data collection, and a lot more careful evaluation and model building than traditional syntax, where one example can be the basis of a new theory, but the results can enrich linguistic theory by revealing the soft constraints at work in language use. This is an area ripe for exploration by the next generation of syntacticians.
Probabilistic Syntax