Search

6 Mar 2008: As for the roughness penalty, we test the penalty set out in Equation 1. ... Several different methods areapplied in tests in Sections 4.3 and 4.4.

15 Aug 2008: Training dataset: about 290hr, 5446spkr; Test dataset: 6hr, 144spkr– Front-end: PLPEnergy1st,2nd,3rd derivatives, HLDA and VTLN used– 16 Gaussian components per state systems; state clustered triphones– 150-Best list

17 Sep 2008: JUD compensation parameters are derived from the joint distribution betweenthe training and test conditions. ... 1289.6 Utterance length mean and standard deviation in TREL-CRL04 test sets.

16 Feb 2008: the speech from the test speaker. These Speaker-Dependent (SD) models are usually. ... the test speaker and the W are now the transcriptions obtained from the SI system.

16 Feb 2008: 716.4 Translation examples for the EPPS Eval05 test set : ASR 1-best and. ... BLEU is defined over all the sentences in the test set i.e.

16 Jan 2008: 137. 8.6 Proportion of development and core test sets rescored as the number of. ... 133. 8.9 Confusable phone-pairs in the TIMIT classification core test set.

8 Jan 2008: 7] A. Martin, “The NIST year 2002 speaker recog-nition evaluation plan,” 2002, available fromhttp://www.nist.gov/speech/tests/spk/2002/doc.

16 Feb 2008: Bitext Word Alignment Word Alignment Results. Bitext Alignment Results. Test: NIST 2001 MT-eval set, 124 sentence pairs w/ manual word alignmentsComparable performance to Model-4 on FBIS training bitextIncreasing max ... GOAL: add phrase pairs to improve

16 Feb 2008: exponential of the cross entropy of the language model on the test set. ... using the posterior distribution with the goal of improving phrase coverage on test.

16 Feb 2008: For each N-best list on the test set, theoracle BLEU hypothesis is computed under the sentence-level BLEUmetric. ... The oracle hypotheses are concatenated over the test set, andthe test-set BLEU score is measured.

26 Mar 2008: This paper investigates using the MBR criterion for unsuper-vised discriminative language model adaptation in test-time for speechrecognition and translation systems. ... Unsupervised test-time discriminative adaptation of mixture languagemodels was

16 Feb 2008: Table: Oracle-best BLEU for EPPS development and test set measured over a1000-best translation list. ... θ̂ = argmaxθ. F(θ). Translation evaluated on a blind test set using the optimized parameters.

9 Jul 2008: 97. 8.2 Singlepass BN Mandarin baseline CER (%) results on the bnmdev06 test set. ... Threshold loss function with β = 0.25 on the dev03 andeval98 test sets.

12 May 2008: D. z. z0. )Tẑ (9). A final test is performed to remove any features located on edges in the image since these willsuffer an ambiguity if used for matching purposes.

19 Mar 2008: Foundations and Trends R inSignal ProcessingVol. 1, No. 3 (2007) 195–304c 2008 M. Gales and S. YoungDOI: 10.1561/2000000004. The Application of Hidden Markov Modelsin Speech Recognition. Mark Gales1 and Steve Young2. 1 Cambridge University

13 Nov 2008: After each test, theestimated axial displacement field was compared with the known ground truth and the averageabsolute point-wise difference d was recorded.