Hidden Markov Models (HMMs) are the most commonly used acoustic model for speech recognition. In HMMs, the probability of successive observations is assumed independent given the ...
This article introduces a new class of constraints for spline variational modeling, which allows more flexible user specification, as a constrained point can ”slide” along a...
Julien Lenoir, Laurent Grisoni, Philippe Meseure, ...
An empirical study has been conducted investigating the relationship between the performance of a generative language model in terms of perplexity and the corresponding informatio...
Leif Azzopardi, Mark Girolami, Keith van Rijsberge...
Parallel computers have the computing power needed to simulate biologically accurate neuronal network models. Partitioning is the process of cutting a model in pieces and assignin...
One of the main obstacles to producing high quality joint models is the lack of jointly annotated data. Joint modeling of multiple natural language processing tasks outperforms si...