Tech Report CS-05-15

Parsing and Disfluency Modeling

Matthew Lease

October 2005


We propose to extend research on parsing and disfluency modeling begun at the 2005 Summer Workshop on Language Engineering at Johns Hopkins University. The proposed research includes: (1) establishing a common definition of fillers across SimpleMDE and tree annotation guidelines, (2) investigating an alternative syntactic representation of fillers for more accurate parsing and more meaningful parser evaluation, (3) improving disfluency modeling by revising and extending existing noisy-channel based approaches, and (4) initiating cross-linguistic study of the interaction between parsing and disfluency modeling.

(complete text in pdf)