Parallelism in Human Parsing

Funding: Deutschen Forschungsgemeinschaft

Project Leader: Prof. Dr. Matthew W. Crocker

Project Start/End: Sept 2004 - August 2006

Project Summary

This project investigates parallelism (the preference for structural similarity) in human sentence processing. In a series of experiments, we will investigate the role that structural, lexical, and metrical information play in establishing parallelism. We will also determine if parallelism is distinct from syntactic priming, and how it interacts with garden path effects in parsing. In the computational part of this project, we will develop incremental, broad-coverage parsing models that can account for our experimental findings. Models will build upon existing probabilistic and neural network models of human parsing, and extend them to deal with parallelism effects and lexical information. The final models will be trained on data from English, German, and Italian.