Overview:

As part of a project looking at how children acquire language, and specifically what features they may use in processing and interpreting language, we will study different psycholinguisticly inspired feature sets for the Semantic Role Labeling task. Data will be restricted to simple sentences from a standard corpora as well as tagged child directed speech. Starting with simple features and assumptions about how children might process speech (such as nouns as arguments) we will attempt to reproduce errors seen at various stages of child development.