Search

13 Jul 2004: NIST,. Benchmark Tests : Rich Transcription (RT).

23 Mar 2004: Woodland: STT Overview. Measuring STT Progress. Current Test Sets : Traditional NIST speech recognition evaluation approach. • ... Progress Test Set : Fixed test set to eliminate test-set variability for trackingprogress towards program goals. •

14 Jun 2004: This was to simulate conditions in the RT-03s evaluation, where there was a temporal gap between the test-audio and the training shows. ... To help minimise the cost of computing the BIC statistics at every point, a faster Hotelling's T test is first

25 May 2004: Test setseval03 6h set from Fisher and Swb2-5 data, 72 conversationsdev04 3h set from Fisher, 36 conversations. ... WER on eval03, fsh2004sub models, unadapted, 2003 trigram. Test with adaptation in P1-P2 system:.

27 May 2004: During training and test adaptation, a global in-terpolation weight was estimated and separate speech and silencetransforms were used for CMLLR. ... Using ST for test adaptation again yieldedabout the same as a GD system with CMLLR.

29 Apr 2004: Test on different systems, architectures and data sets. • Allow more than 2 inputs. ... Use dual-phone recogniser to remove long silences. • Use LIMSI-style segmentation/clustering algorithm [ NB - parameters havebeen chosen on the test data].

18 Jun 2004: During training and test adaptation, aglobal interpolation weight was estimated and separate speech andsilence transforms were used for CMLLR. ... Using ST for test adaptation again yieldedabout the same as a GD system with CMLLR.

27 May 2004: Experimental Setup• Unadapted results• Conversational telephone speech – English (CTS):. – Training dataset: h5etrain03 (296hr)– Test dataset: dev01sub (3hr) & eval03 (6hr)– CMN, CVN and VTLN are used– Basis vectors/matrices: ML trained

25 May 2004: 8. REFERENCES. [1] NIST, “The Rich Transcription Spring 2003 (RT-03S)Evaluation Plan, version 4,” http://www.nist.gov/speech/tests/rt/rt2003/spring/docs/rt03-spring-eval-plan-v4.pdf , ... 2] NIST, “Benchmark Tests : Rich Transcription (RT),”http:/

27 May 2004: Training and Test Data Sets. • Acoustic training dataTDT4 Oct 2000 - Jan 2001. – ... Corrected closed-captions subset– 370 shows, 250h recorded broadcasts. • Test setsdev03 17th Jan 2001 - 31 Jan 2001 (6 shows, 3h data)eval03 Feb 2001 (6 shows,

25 May 2004: κ is a scalingfactor and can be adjusted to improve test-set performance. ... Table 3. Comparison of WER for MPE trained HLDA and EM-LLT models on thedev01sub test set.

29 Apr 2004: There are some small gains using ensemble techniques, but the gains are notconsistent across the dev03f and eval03f test sets. ... The PFM combined with LDCDB98 SULMs can give some small gains, but thepatterns are not consistent across the dev03f and

27 May 2004: GI unadapted single pass decoding system. Table 2 gives WER on the test sets using the5xRT P1-P2system. ... thedev03 and eval03 test sets respectively.These results were even slightly better than the full 10xRT CU-.

23 Mar 2004: The warp-factors for test-data are estimated using the VTLN-MLE model.dev03 eval03. ... System Training Test EstimationAdaptation Adaptation MLE MPE. GI —— 33.4 30.4.

25 Mar 2004: http://www.nist.gov/speech/tests/rt/rt2003/spring/docs/ref-cookbook-v2_4.pdf , 17th March 2003. ... 6] NIST, “Benchmark Tests : Rich Transcription (RT),”http://www.nist.gov/speech/tests/rt/. [7] S.E. Tranter, K.

27 May 2004: The evaluation test set used in the 2003 DARPA/NIST RichTranscription evaluation (eval03) contains data from the LDC Fishercollection1 and from Switchboard II phase 5. ... Vocal TractLength Normalisation (VTLN) was used in training and test bywarping the

22 Mar 2004: SU-Detection System for RT-03F Evaluation. Results for the dev03 (3hrs) and eval03 (3hrs) test sets:.

28 Jul 2004: 5 Evaluation. To test our global localisation system, we constructed a database of views by photograph-ing all the buildings in the main shopping street in Cambridge’s city centre, as

22 May 2004: The training data set comprised 3990 utterances. All1200 test utterances (feb89, oct89, feb91, sep92), tst, and a 300utterance subset of the training data, trn, with a simple word-pair grammar, were

26 Jan 2004: The test set used wasthe subset of the 1997 Hub5 evaluation set used in [9]. ... The test set error rates for all the configurations describedabove are depicted in Table 3.