Language models (Steve) - Model M

Zweig and Chang, Personalising Model M, Interspeech 2011

Chen et al, Scaling shrinkage-based LMs, Extended version of ASRU-2009

Sethy et al, Distributed training of Model M, ICASSP 2011

Brown et al, Class-based n-grams, Computational Linguistics 1992

ASR with a KL-HMM Magimai-Doss et al (Peter)

Diarization (Mark)

Information Bottleneck Features for HMM/GMM Speaker Diarization of Meetings

IS110316.PDF

Integrated Online Speaker Clustering and Adaptation

IS110311.PDF

Noisy ASR (Liang):

Noise model estimation: (zhao2010comparative.pdf)

Joint Uncertainty Decoding: (xu2009comarison.pdf)

For more background, see Mark Gales' review (gales_noise_review10.pdf)

Deep architectures 1 (Pawel, Steve)

Acoustic Modelling using Deep Belief Networks (mohamed_hinton2011.pdf)

Optional reading for broader background on deep networks:

Hinton, et al. A fast learning algorithm for deep belief networks (hinton2006_deep.pdf)

Yoshua Bengio, Learning Deep Architectures for AI (bengio2009_deep_ai.pdf)

Deep architectures 2 (Pawel, Peter)

LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION WITH CONTEXT-DEPENDENT DBN-HMMS DBN4LVCSR-TransASLP.pdf

ASR of Lombard speech (Cassia)

Unsupervised equalization of Lombard effect for speech recognition in noisy adverse environments LE_IEEETrans.pdf

UT-SCOPE: Towards LVCSR under Lombard effect induced by varying types and levels of noisy background LE_ICASSP11.pdf

NB change of time - probably 3pm.

CSTR's recognition setup (Fergus)

Background reading:

IEEE Transactions paper (in draft) on the AMIDA RT07 and RT09 systems: Hain et al., Transcribing meetings with the AMIDA systems

We won't be going through the paper in detail, but there will be some slides on the present state of the system and how to obtain and use it.

Here are the slides: Listen-2011-12-07.ppt