Search

2 Nov 2000: 2 Finding the TopicsA mixture of broadcast news and newswire text was usedas training data for the topic model, with 144 million wordsof Broadcast News text and 25 million words of ... 3 Building a Language ModelA word 4-gram language model was built

20 Apr 2000: Also a windowing systemis beingdevelopedthatallows automaticcontentdependentsegmentationof news broadcasts. Acknowledgements. This work is in partfundedby anEPSRCgrantreferenceGR/L49611. ... In Proc. DARPA Broadcast News Workshop,

10 May 2000: We processed this data to automatically detect andremove commercials [2]. Transcription used a simplifiedversion of the HTK broadcast news system which corre-sponds to the “first-pass” recognition system describedin [2]. ... Thespeech recogniser uses

10 Apr 2000: A full-scale implementation is then de-scribed and tested on data from the 1996 Broadcast News database. ... 2. 1.2 DataThe data used throughout this project was from the 1996 Broadcast News database.

5 Oct 2000: A 4-gram language model was trained on the transcripts of theacoustic training data and additional broadcast news transcriptions.The word 4-gram was smoothed with a class-based trigram that

6 Nov 2000: It was reported in [32] that FD didn’t improveerror rates over MLE trained models for a broadcast newsrecognition task. ... The system uses word-based N-gramLMs estimated from an interpolation of Hub5 acoustic train-ing transcriptions and Broadcast