Search

9 Aug 2005: British Machine Vision Conference 4. Figure 2: A selection of the test images. ... test and Jef-frey divergence (identical) (C&J), and the EarthMover distance (EM).

16 Mar 2005: Further details andan implementation of the K-S test can be found in [27]. ... 10. 20. 30. 40. 50threshold for KS test statistics. threshold for KS test statistics.

10 Jan 2005: nist.gov/speech/tests/rt/rt2003/spring/docs/rt03-spring-eval-plan-v4.pdf , 25th February 2003. ... 3] NIST, “Fall 2004 Rich Transcription (RT-04F) EvaluationPlan,” http://www.nist.gov/speech/tests/rt/rt2004/fall/docs/rt04f-eval-plan-v14.pdf , 30th

9 Aug 2005: Vocal tract length nor-malisation (VTLN) was applied in both training and test. ... Similar im-provements have been found on other test sets. 7. FULL VARIANCE TRANSFORMS.

9 Aug 2005: They are obtained by optimising the inser-tion penalty and another parameter, the silence probability incre-ment, for each test condition. ... Lynx noisefrom the NOISEX-92 database was added to the test sets at 18dBand 12dB.

9 Aug 2005: where4. is retained inE G. when the test succeeds. In practice the value of is fixed (e.g. ... Training- and test-sets were created by splitting thematerial evenly across these topics in the ratio 95:5, resulting in a vocabulary size of 41097 words.

15 Feb 2005: Presentation Overview. • RT03 Broadcast News System Review• Training & Test Data• Improved Acoustic Model Building. – ... Test set # Shows Size Period. dev03 6 3hrs Jan01eval03 6 3hrs Feb01dev04 6 3hrs Jan01dev04f 6 3hrs Nov03.

15 Feb 2005: Training and Test Data for CTS. The following sets of CTS training data were used:. ... Training and Test Data for BN. The following sets of BN training data were used:.

19 May 2005: This system wasevaluated on both eval03 and dev04 test sets and comparedwith the SATDIAGC system. ... Adaptation experimentswere conducted based on three 3-hour test sets: eval03, dev04and dev04f.

26 Mar 2005: SID stage - Adding Eval Data to UBM. • New UBMs, BE and BMAPE, were created using the whole test data set.• System gender labels were used. ... no real cheating but against eval rules).• Using just the target show (rather than whole test set) did

9 Aug 2005: Particular importance must be given to the effect of words en-countered in the test data which have not been seen in the trainingdata. ... We also used 3 hours of test data from the NIST 1998 Hub-4BN benchmark tests.

12 Apr 2005: Acoustic Model Test Data:– eval03 6 hours (3 hours Switchboard2 Phase 5, 3 hours Fisher)– dev04 3 hours Fisher data. • ... Experimental Setup. • Acoustic model data sets:– Training data: 76 hours h5etrain03sub & 296 hours h5etrain03– Test data

9 Aug 2005: and kill-tests respectively, are used to decide upon the strengthof the correlation. ... The kill-test succeeds when 8 X , and the following figure illustrates these conditions.

10 Jan 2005: The transcription is either known in training orobtained by an initial, non-VTLN, decoding pass for test data.If and are the original and transformed feature vectorsrespectively then the log-likelihood ... The systems were eval-uated on two 3 hour test

8 Aug 2005: These meth-ods have achieved very good accuracy on a small number ofcontrolled test sets. ... We therefore useDKL(p(0)||p(i)) as a “distance measure” between trainingand test sets.

9 Aug 2005: At each update a t-test isperformed, determining to a desired level of confidence whetherthe measured mean is lower than would be expected were the can-didates to occur independently. ... The following table gives a brief description ofeach corpus, where

25 Mar 2005: 6. REFERENCES. [1] NIST, “Benchmark Tests : Rich Transcription (RT),” http://www.nist.gov/speech/tests/rt/. ... 2] NIST, “The Rich Transcription Spring2003 (RT-03S) Evaluation Plan, version 4,”http://www.nist.gov/speech/tests/rt/rt2003/spring/docs

19 May 2005: Among all the adaptivelytrained systems, the SATSPAM outperformed all the other sys-tems on both test sets. ... 3. Three BN test sets were used, each of them contains six 30 minutesbroadcast news shows.

9 Aug 2005: Steps 6, 3, 4 and 5 are then carried out to estimateand test a pronunciation model. ... in Step 5 (Test) PER WER PER WERNone (Dictionary) 49.1% 49.1% 49.5% 58.9%Tree Pron.

26 Sep 2005: Here. P (šn|xt, M̌) P (šn|yt, M̌) (10)the model M̌ is now matched to the test condition rather. ... All results are quotedaverage of three of the four available test sets, Feb’89,89 and Feb’91.