Paper: Is Sentence Compression an NLG task?

ACL ID W09-0604
Title Is Sentence Compression an NLG task?
Venue European Workshop on Natural Language Generation
Year 2009

Data-driven approaches to sentence com- pression define the task as dropping any subset of words from the input sentence while retaining important information and grammaticality. We show that only 16% of the observed compressed sentences in the domain of subtitling can be accounted for in this way. We argue that part of this is due to evaluation issues and estimate that a deletion model is in fact compat- ible with approximately 55% of the ob- served data. We analyse the remaining problems and conclude that in those cases word order changes and paraphrasing are crucial, and argue for more elaborate sen- tence compression models which build on NLG work.