Sentence-Internal Prosody Does not Help Parsing the Way Punctuation Does

This paper investigates the usefulness of sentence-internal prosodic cues in syntactic parsing of transcribed speech. Intuitively, prosodic cues would seem to provide much the same information in speech as punctuation does in text, so we tried to incorporate them into our parser in much the same way as punctuation is. We compared the accuracy of a statistical parser on the LDC Switchboard treebank corpus of transcribed sentence-segmented speech using various combinations of punctuation and sentence-internal prosodic information (duration, pausing, and f0 cues). With no prosodic or punctuation information the parser’s accuracy (as measured by F-score) is 86.9%, and adding punctuation increases its F-score to 88.2%. However, all of the ways we have tried of adding prosodic information decrease the parser’s F-score to between 84.8% to 86.8%, depending on exactly which prosodic information is added. This suggests that for sentence-internal prosodic information to improve speech transcript parsing, either different prosodic cues will have to used or they will have be exploited in the parser in a way different to that used currently.