logo

MATE Deliverable D1.1

Supported Coding Schemes

SWBD-DAMSL
(University of Colorado)
 

Coding book:
http://stripe.Colorado.EDU/~jurafsky/manual.august1.html
Authors: Dan Jurafsky, Liz Shriberg, Debra Biasca
Title: Switchboard SWBD-DAMSL, Shallow-Discourse-Function Annotation; Coders Manual, Draft 13

Number of annotators:
9 (linguists)

Number of dialogues annotated:
1155 conversations / over 250 000 utterances / 1.4 million words

Evaluation of scheme:
0.80 < k < 0.84

Underlying task:
telephone conversation between strangers

List of phenomena annotated:


Examples:
 

ad A63  utt2:  {C and } think [what, + what's ] is going to be like for [ youngest, ] + [an + ] my oldest ] son, when he goes to school. 
qh A.63 utt3:  What's going to happen?  /
sd A.63 utt4:  {E I mean } [ I, + I'm ] afraid for him to go. / 

Mark-up language:
Variant of DAMSL

Existence of annotation tool:
None. Utterances are hand-labeled.

Usability:
Used for training stochastic discourse grammars so as to build better Language Models (LM) for Automatic  Speech Recognition (ASR) of Switchboard. A variant of SWBD-DAMSL is CLARIFY currently developed at the Carnegie Mellon University (Lori_Levin@alexis.boltz.cs.cmu.edu).

Contact person:
Daniel Jurafsky (jurafski@colorado.edu)

Last Modification: 27.8.1998 by Marion Klein