Search
Search Funnelback University
151 -
200 of
1,000
search results for katalk:PC53 24 / |u:mi.eng.cam.ac.uk
where 0
match all words and 1,000
match some words.
Results that match 1 of 2 words
-
Cross-Lingual Spoken Language Understanding from Unaligned Data…
mi.eng.cam.ac.uk/~sjy/papers/lemy10.pdf20 Feb 2018: on Mancorpora)a 91.40 90.17 90.24. 90.20Auto (Google MT) 90.81 90.77 87.72 89.223. ... 24, no. 2, pp. 150–174, April 2010. [8] P. Koehn, H. -
acl2010.dvi
mi.eng.cam.ac.uk/~sjy/papers/gjkm10.pdf20 Feb 2018: Computer Speech and Language, 24(2):150–174. -
tech.dvi
mi.eng.cam.ac.uk/~sjy/papers/bghk13.pdf20 Feb 2018: 24, pp. 562–588, 2010. [3] G. Aist, J. Allen, E. Campana, C. ... 24] M. Henderson, M. Gašić, B. Thomson, P. Tsiakoulis, K.Yu,and S. -
sigdial11_sdc10-Feb27-V2
mi.eng.cam.ac.uk/~sjy/papers/bbch11.pdf20 Feb 2018: 6% 24.6% 14.7% 9.6%. ... Length (s) Turns/call Words/turn. SYS1 control 155 18.29 2.87 (2.84) SYS1 live 111 16.24 2.15 (1.03) SYS2 control 147 17.57 1.63 -
Incremental on-line adaptation of POMDP-based dialogue managers…
mi.eng.cam.ac.uk/~sjy/papers/gktb14.pdf20 Feb 2018: 24, no. 2, pp.150–174, 2010. [5] B. Thomson and S. Young, “Bayesian update of dialogue state:A POMDP framework for spoken dialogue systems,” ComputerSpeech and Language, vol. ... 24, no. 4, pp. 562–588, 2010. [6] M. Gašić, C. Breslin, M. -
1 Structured Log Linear Models for Noise RobustSpeech Recognition ...
mi.eng.cam.ac.uk/~mjfg/zhang10.pdf8 Sep 2010: Speech Lang.,vol. 24, no. 4, pp. 648–662, 2010. [8] B. Taskar, “Learning structured prediction models: a large marginapproach,” Ph.D. -
Reward Shaping with Recurrent Neural Networks for Speeding upOn-Line…
mi.eng.cam.ac.uk/~sjy/papers/svgm15.pdf20 Feb 2018: Computer Speech and Language,24:562–588. Jason D. Williams and Steve Young. 2007. -
Dialogue Context Sensitive Speech Synthesis using Factorized Decision …
mi.eng.cam.ac.uk/~sjy/papers/tgbh14.pdf20 Feb 2018: In mostcases, a data-driven approach is followed, either by detect-ing/annotating emphasized words in existing corpora [23, 10] orby collecting speech corpora specifically designed for emphasismodeling [24]. ... Appointment Booking Task -
Uncertainty management for on-line optimisation of a…
mi.eng.cam.ac.uk/~sjy/papers/dgcg11.pdf20 Feb 2018: 24, no. 2, pp. 150–174,2010. [6] W. Eckert, E. Levin, and R. ... 9] O. Pietquin, M. Geist, S. Chandramohan, and H. Frezza-Buet,“Sample-Efficient Batch Reinforcement Learning for DialogueManagement Optimization,” ACM Transactions on Speech -
Combining I-vector Representation and Structured Neural Networks for…
mi.eng.cam.ac.uk/~mjfg/icassp16_wu.pdf5 Apr 2016: Speech andSignal Processing, ICASSP 2015, South Brisbane, Queens-land, Australia, April 19-24, 2015, 2015, pp. ... IEEE, 2015, pp. 4315–4319. [24] Mark JF Gales, “Cluster adaptive training of hidden markovmodels,” Speech and Audio Processing, IEEE -
Learning Domain-Independent Dialogue Policies via…
mi.eng.cam.ac.uk/~sjy/papers/wsws15.pdf20 Feb 2018: Computer Speech andLanguage, 24(4):562–588. Zhuoran Wang and Oliver Lemon. 2013. A simpleand generic belief tracking mechanism for the Dia-log State Tracking Challenge: On the believabilityof observed information. -
Learning from Real Users: Rating Dialogue Success with Neural ...
mi.eng.cam.ac.uk/~sjy/papers/svgk15.pdf20 Feb 2018: 24, pp. 562–588, 2010. [20] M. Lukoeviius and H. Jaeger, “Reservoir computing approachesto recurrent neural network training,” Computer Science Review,vol. ... abs/1412.2306, 2014. [24] G. Mesnil, Y. Dauphin, K. Yao, Y. Bengio, L. -
Evaluation of Statistical POMDP-basedDialogue Systems in Noisy…
mi.eng.cam.ac.uk/~sjy/papers/ybgh14.pdf20 Feb 2018: When errors are correlated belief tracking is less accurate be-cause it tends to over-estimate alternatives in the N-best list[24]. ... 24, no. 4, pp. 562–588, 2010. 17. T. Minka, “Expectation Propagation for Approximate Bayesian Inference,” in -
LEARNING BETWEEN DIFFERENT TEACHER AND STUDENT MODELS IN ASR ...
mi.eng.cam.ac.uk/~mjfg/ALTA/ASRU2019_TS.pdf20 Dec 2019: The derivatives of the per-frame. observation log-likelihoods with respects to the parameters are [24]. ... Work in [24] suggests several methods to improve gra-dient descent training of a GMM. -
is-05-hvs6_final
mi.eng.cam.ac.uk/~sjy/papers/seyo05.pdf20 Feb 2018: 52 class n-gram 26.3 25.0 24.9 HVS_52 21.7 20.4 20.1. Table 1: Perplexity for models of varied stack depths trained for 250 iterations. -
poyosp08
mi.eng.cam.ac.uk/~sjy/papers/dpyo08.pdf20 Feb 2018: T - R 0.54 0.22 0.24. R - O 0.52 0.31 0.17. -
yokou
mi.eng.cam.ac.uk/~sjy/papers/toyo09.pdf20 Feb 2018: where. x(d)q = 2. “c(d)q. Dc(d)q. E”(v(cq) v(c)) p(d)v. (24). 4.3. -
Deep Learning for Speech Processing - An NST Perspective
mi.eng.cam.ac.uk/~mjfg/NST_2016.pdf29 Sep 2016: 24 of 67. S2S: Generative Models [5, 6]. • Consider two sequences L T: input: x1:T = {x1, x2,. , ... 47 of 67. ASR: Sequence Training [24]. • Cross-Entropy using fixed alignment standard criterion (RNN). -
SYSTEM COMBINATION WITH LOG-LINEAR MODELS J. Yang, C. Zhang, ...
mi.eng.cam.ac.uk/~mjfg/yang_ICASSP16.pdf12 Jul 2016: When the segment level features are used, the log-linear modelparameters η̂ could be considered as phone dependent acousticmodel scales [24]. ... In joint decoding, 2% relative WER performance gain wasachieved over the hybrid system, from 11.24% to -
DISTRIBUTED DIALOGUE POLICIES FOR MULTI-DOMAIN STATISTICAL…
mi.eng.cam.ac.uk/~sjy/papers/gkty15a.pdf20 Feb 2018: 24, no. 4, pp. 562–588, 2010. [13] M Gašić, C Breslin, M Henderson, D Kim, M Szummer,B Thomson, P Tsiakoulis, and S Young, “POMDP-based dia-logue manager adaptation -
Online_ASRU11.dvi
mi.eng.cam.ac.uk/~sjy/papers/gjty11.pdf20 Feb 2018: 24, no. 2, pp. 150–174, 2010. [9] B. Thomson and S. ... 24, no. 4, pp. 562–588, 2010. [10] M. Gǎsić, S. Keizer, F. -
Part IA Computing CourseLent Term Software Design Exercise Roberto ...
mi.eng.cam.ac.uk/~cipolla/resource/lent.pdf12 Apr 2022: default:cout << "Trading error: trading system failure." << endl;exit(-1);. }}. 24. -
JOINT MODELLING OF VOICING LABEL AND CONTINUOUS F0 FOR ...
mi.eng.cam.ac.uk/~sjy/papers/yuyo11a.pdf20 Feb 2018: Mixed excitation using STRAIGHT was employed [12].The speech features used were 24 Mel-Cepstral spectral coefficients,the logarithm of F0, and aperiodic components in five frequencybands (0 to 1, 1 -
System Combination with Log-linear Models
mi.eng.cam.ac.uk/~mjfg/icassp16_yang.pdf5 Apr 2016: When the segment level features are used, the log-linear modelparameters η̂ could be considered as phone dependent acousticmodel scales [24]. ... In joint decoding, 2% relative WER performance gain wasachieved over the hybrid system, from 11.24% to -
Infinite Support Vector Machines in Speech Recognition Jingzhou Yang, …
mi.eng.cam.ac.uk/~mjfg/yang_is2013.pdf13 Jun 2013: Better performance couldbe achieved by gradually increasing C. Equation (16) is also known as the training criterion ofthe structural SVM [23, 24]. ... ACM, 2004. [24] Shi-Xiong Zhang and Mark Gales, “Structured SVMs for auto-matic speech -
Domain-independent User Satisfaction Reward Estimation forDialogue…
mi.eng.cam.ac.uk/~sjy/papers/ubcm17.pdf20 Feb 2018: Sec. 2. the reward [20, 21, 22, 23] by using the PARADISE frame-work [24]. ... 2356–2361. [24] M. Walker, D. J. Litman, C. A. Kamm, and A. -
slides.dvi
mi.eng.cam.ac.uk/~mjfg/Bilbao14/talk.pdf25 Jun 2014: Cambridge University. Engineering DepartmenteNTERFACE June 2014 24. Controllable and Adaptable Statistical Parametric Speech Synthesis Systems. ... Integrated Expressive Speech Training [24]. Training. Expressive State. Prediction. ExtractionAcoustic -
Low-Resource Speech Recognition and Keyword-Spotting
mi.eng.cam.ac.uk/~mjfg/SPECOM_2017.pdf29 Nov 2017: 23/63. Stimulated Systems. /ey//em/. /sil/. /sh/. /ow/ /ay/. 24/63. Stimulated Network Training. • -
Tsiakoulis_Pirros_1293
mi.eng.cam.ac.uk/~sjy/papers/tghp12.pdf20 Feb 2018: Proc. ICASSP, Taipei, Taiwan. Thomson, B. & Young, S. (2010)“Bayesian Update of Dialogue State: A POMDP framework for spoken dialogue systems.” Computer Speech and Language 24(4):562-588. -
Natural Belief-Critic: a reinforcement algorithm for parameter…
mi.eng.cam.ac.uk/~sjy/papers/jtkm10.pdf20 Feb 2018: 24, no. 4, pp. 562 – 588, 2010. [2] R. Sutton and A. -
main.dvi
mi.eng.cam.ac.uk/~sjy/papers/youn0720 Feb 2018: Hence, an itera-tive algorithm can be implemented which repeatedlyscans through the vocabulary, testing each word tosee if moving it to some other class would increasethe likelihood [24]. ... Thed p nuisance dimensions are modelled by a -
paper.dvi
mi.eng.cam.ac.uk/~mjfg/yw293_ASRU11.pdf19 Jan 2012: and. J(m)xδ =. g. xtδ. µ(m)xe ,µl,µn. (24). Thus the model parameters are compensated by. ... Forexample, using the initial noise estimate, RVTSJ performancevaried from 27.5% to 31.7%, while the performance of MLestimated noise only varied from 24.3% -
MULTI-DOMAIN DIALOGUE SUCCESS CLASSIFIERS FOR POLICY TRAINING David…
mi.eng.cam.ac.uk/~sjy/papers/vsgm15.pdf20 Feb 2018: 6. RELATED WORK. In machine learning in general much research has looked atadaptation of statistical models [21, 22, 23] however researchinto adaptation of SDS components to new domains [24, 25,26, ... 24, pp.562–588, 2010. [16] Milica Gašić and -
SSVM_LVCSR_ASRU11.dvi
mi.eng.cam.ac.uk/~mjfg/sxz20_ASRU11.pdf19 Jan 2012: 23)), finds the most violated constraint (Eq. (24)), andadds it to the working set. ... Parallelingthe loop for Eq. (24) will lead to a substantial speed-up in thenumber of threads. -
Investigation of back-off based interpolation between Recurrent…
mi.eng.cam.ac.uk/~mjfg/asru15-chen.pdf11 Mar 2016: In order to solve this problem, recentlythere has increasing research interest in deriving efficient paralleltraining algorithms for RNNLMs [22, 23, 24, 25]. ... 9.1 2.9 117.85GRNN 5472.1 170.0 24.3 7.9 2.4 117.6. -
Exploiting Sentence and Context Representations in Deep Neural…
mi.eng.cam.ac.uk/~sjy/papers/rgms16.pdf20 Feb 2018: F1 ICE. Slot4 95.29% 90.89% 95.72% 93.24% 0.478 89.92% 74.73% 61.56% 67.51% 0.743. ... Task - - - - - 97.12% 83.24% 64.93% 72.95% 0.175. -
draft21.dvi
mi.eng.cam.ac.uk/~mjfg/ASRU13.pdf7 Nov 2013: procedure described in [24]. Unilingual and multilingual AMs were. each built from a flat start. ... 24] J. Park et al., “The Efficient Incorporation of MLP Fea-. -
THE USE OF DISCRIMINATIVE BELIEF TRACKINGIN POMDP-BASED DIALOGUE…
mi.eng.cam.ac.uk/~sjy/papers/khgt14.pdf20 Feb 2018: 24,no. 4, pp. 562–588, Oct. 2010. [7] Jost Schatzmann, Statistical user and error modellingfor spoken dialogue systems, Ph.D. -
Multi-domain Neural Network Language Generation forSpoken Dialogue…
mi.eng.cam.ac.uk/~sjy/papers/wgmr16.pdf20 Feb 2018: scr-10% 2.24 2.03 2.00 1.92. p <0.05, p <0.005Table 2: Human evaluation for utterance quality intwo domains. -
paper.dvi
mi.eng.cam.ac.uk/~mjfg/sxz20_inter11.pdf19 Jan 2012: SpeechLang., vol. 24, no. 4, pp. 648–662, 2010. [6] B. Taskar, “Learning structured prediction models: a large marginapproach,” Ph.D. -
DEVELOPMENT OF THE 2003 CU-HTK CONVERSATIONAL TELEPHONE…
mi.eng.cam.ac.uk/reports/svr-ftp/evermann_icassp2004.pdf27 May 2004: purpose WER. P1 supervision for VTLN 34.2P2 supervision for MLLR 28.4P3 lattice generation 24.8. ... System (P4) A B C DSAT HLDA SPron non-HLDA23.0 23.6 23.4 24.8. -
3_2_ransac
mi.eng.cam.ac.uk/~cipolla/lectures/4F12/Slides/4F12-ImageStitching.pdf27 Oct 2020: x̃2 = [ R2 | 0 ] X̃ = R2X (A.24). -
Reward Estimation for Dialogue Policy Optimisation Pei-Hao Su, Milica …
mi.eng.cam.ac.uk/~sjy/papers/sugy18.pdf20 Feb 2018: In this section, a Gaussianprocess-based reward estimator is described which uses active learning tolimit intrusive requests for feedback and a noise model to mitigate the effectsof inaccurate feedback [24]. ... 24. Figure 13: The number of times each -
lect1.dvi
mi.eng.cam.ac.uk/~mjfg/local/4F10/lect1.pdf10 Nov 2015: for minimum error with generative models. 24 Engineering Part IIB: Module 4F10 Statistical Pattern Processing. -
SYSTEM COMBINATION AND SCORE NORMALIZATION FOR SPOKEN TERM DETECTION
mi.eng.cam.ac.uk/~mjfg/ICASSP13_ibm1.pdf13 Jun 2013: Ney LM with optimized discounting parameters [24] usinga modified version of the RWTH open source decoder [25]; and (6)DBN, a deep belief network hybrid model [26, 27] with discrimi-native ... Ney, “Posterior-scaled mpe: Novel discriminative -
Unsupervised Language Model Adaptation for Mandarin…
mi.eng.cam.ac.uk/reports/svr-ftp/mrva_icslp06.pdf20 Jan 2007: Test set baseline N-gram adaptfixed weights dynamic weights. dev05bcm (BC) 25.6 24.5eval04 (BN) 14.7 14.8dev04f (BN) 6.4 6.5. ... P3 27.4 25.6 24.5 24.3 24.3. Table 3: P2, P3 stage dev05bcm CERs. -
paper.dvi
mi.eng.cam.ac.uk/reports/svr-ftp/liu_icassp2004.pdf29 May 2004: Gauss 24.0 20.7 20.7 17.7 16.0WER (%) 35.3 35.1 35.2 35.3 35.5. ... The 16 component systems was then iteratively split until thenumber of components was 24. -
Statistical User Simulation with a Hidden Agenda Jost Schatzmann ...
mi.eng.cam.ac.uk/~sjy/papers/scty07.pdf20 Feb 2018: U| 103 and |M| 103. (24). Goals are composed ofNC constraints taken from theset of constraintsC, andNR requests taken from the setof requestsR. -
eps.dis.dur.testa.eps
mi.eng.cam.ac.uk/~mjfg/gales_ASRU09.pdf14 Sep 2010: Using 17 pairs, about 24% of thetotal number of pairs, 92% of the WER improvement usingthe 1-v-1 system over the VTS baseline was achieved. -
Quantitative Evaluation of User Simulation Techniquesfor Spoken…
mi.eng.cam.ac.uk/~sjy/papers/scgy05.pdf20 Feb 2018: The Pietquin model. Train TestPrecision Recall Precision Recall. BIG 19.74 24.11 17.83 21.66LEV 43.11 35.07 37.98 31.57PTQ 45.00 36.35 40.16
Refine your results
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.