Skip to search formSkip to main contentSkip to account menu

Purely Sequence-Trained Neural Networks for ASR Based on Lattice-Free MMI

TLDR
A method to perform sequencediscriminative training of neural network acoustic models without the need for frame-level cross-entropy pre-training is described, using the lattice-free version of the maximum mutual information (MMI) criterion: LF-MMI.
View via Publisher (opens in a new tab)

Semi-Supervised Training of Acoustic Models Using Lattice-Free MMI

TLDR
Various extensions to standard LF-MMI training are described to allow the use as supervision of lattices obtained via decoding of unsupervised data and different methods for splitting the lattices and incorporating frame tolerances into the supervision FST are investigated.
View on IEEE (opens in a new tab)

Diarization is Hard: Some Experiences and Lessons Learned for the JHU Team in the Inaugural DIHARD Challenge

TLDR
Several key aspects of currently state-of-the-art diarization methods, such as training data selection, signal bandwidth for feature extraction, representations of speech segments (i-vector versus x-vector), and domainadaptive processing are explored.
View via Publisher (opens in a new tab)

JHU ASpIRE system: Robust LVCSR with TDNNS, iVector adaptation and RNN-LMS

TLDR
This paper tackles the problem of reverberant speech recognition using 5500 hours of simulated reverberant data using time-delay neural network (TDNN) architecture, which is capable of tackling long-term interactions between speech and corrupting sources in reverberant environments.
View on IEEE (opens in a new tab)

Acoustic Modelling from the Signal Domain Using CNNs

TLDR
The resulting ‘direct-fromsignal’ network is competitive with state of the art networks based on conventional features with iVector adaptation and, unlike some previous work on learned feature extractors, the objective function converges as fast as for a network based on traditional features.
View via Publisher (opens in a new tab)

Investigation of transfer learning for ASR using LF-MMI trained neural networks

TLDR
This paper investigates different ways to utilize out-of-domain data to improve ASR models based on Lattice-free MMI (LF-MMI), and experiments with multi-task training using a network with shared hidden layers and various ways of adapting previously trained models to a new domain.
View on IEEE (opens in a new tab)

An Exploration of Dropout with LSTMs

TLDR
This paper describes extensive experiments in which the best way to combine dropout with LSTMs– specifically, projected LST Ms (LSTMP) is investigated, giving consistent improvements in WER across a range of datasets, including Switchboard, TED-LIUM and AMI.
View via Publisher (opens in a new tab)

Far-Field ASR Without Parallel Data

TLDR
This paper explores methods to reduce the performance gap between far-field ASR systems trained with alignments from distant microphone data and those trained with alignment from parallel close-talk microphone data, and identifies prominent causes of alignment errors in AMI data.
View via Publisher (opens in a new tab)

Characterizing Performance of Speaker Diarization Systems on Far-Field Speech Using Standard Methods

TLDR
This work evaluates the performance of an ivector/PLDA-based diarization system on the AMI Meeting Corpus, comparing performance on near- field, far-field, and signal-enhanced conditions.
View on IEEE (opens in a new tab)

Semi-supervised maximum mutual information training of deep neural network acoustic models

TLDR
It is shown that if the supervision transcripts are not known, the natural analogue of MMI is to minimize the conditional entropy of the lattice of possible transcripts of the data, equivalent to the weighted average of M MI criterion over different reference transcripts, taking those reference transcripts and their weighting from the lattICE itself.
View Paper (opens in a new tab)
By clicking accept or continuing to use the site, you agree to the terms outlined in our Privacy Policy (opens in a new tab), Terms of Service (opens in a new tab), and Dataset License (opens in a new tab)