Paper: Summarizing Speech Without Text Using Hidden Markov Models

ACL ID N06-2023
Title Summarizing Speech Without Text Using Hidden Markov Models
Venue Human Language Technologies
Session Short Paper
Year 2006
Authors

We present a method for summarizing speech documents without using any type of transcript/text in a Hidden Markov Model framework. The hidden variables or states in the model represent whether a sentence is to be included in a sum- mary or not, and the acoustic/prosodic fea- tures are the observation vectors. The model predicts the optimal sequence of segments that best summarize the docu- ment. We evaluate our method by compar- ing the predicted summary with one gen- erated by a human summarizer. Our re- sults indicate that we can generate ’good’ summaries even when using only acous- tic/prosodic information, which points to- ward the possibility of text-independent summarization for spoken documents.