Language Learning in Large Parameter Spaces

The existence of parameters has been proposed in models of linguistic theory to account for differences among natural languages. In addition to the problem of defining parameters, we have the problem of a child’s acquisition of the settings of these parameters. Several algorithms have been proposed to describe how a child learns the parameter settings for her target adult language, but these algorithms need to be analyzed in greater depth. We used an implentation of one proposed algorithm of parameter setting to study its predictions in a more realistic setting. It was necessary to implement this algorithm for large parameter spaces in order to see that its problems were serious and that the problem of parameter setting cannot easily be solved.

Authors

• Karen T. Kohl MIT Artificial Intelligence Laboratory

Keywords

No keywords are indexed for this paper.