Kaldi scales down the acoustic cost: cost = am_cost * am_scale + lm_cost (am_scale...
How are you combing the systems? On 3 November 2014 17:29, Bhargav Pulugundla bpulugundla@users.sf.net...
ERROR! The markdown supplied could not be parsed correctly. Did you forget to surround...
It could be an IO bottleneck, or the memory of the machine is exhausted and the processes...
Does it work if you use the default OSX compiler(clang)? On 15 July 2014 07:40, enzyme156...
Which type of acoustic model? On 19 June 2014 08:12, rayava rayava@users.sf.net wrote:...
You could try increasing the beam or max tokens for the biglm decoder. On Friday,...
Finally, I'd love any insight on how best to use my transcript to build the language...
try featbin/paste-feats On 12 May 2014 17:58, sprieto sprieto@users.sf.net wrote:...
Your question is a little bit hard to follow, but you can build simple transducers...
sandbox/paul_dixon: Copy of Nickolay's TED scripts
Lattice to far converter to allow for interop b...
By WFST approach you mean the Gaussian computations or the entire decoding process?...
sandbox/paul_dixon:Updated to latest trunk
sandbox/paul_dixon:Updated to latest trunk and ...
Seems like something is not right before running the fstrmepsilon command. Trying...
You have to do the conversion to LogArc after extracting the shortest path. On 15...
Probably the quickest is to use ArcMap from arc-map.h to convert the StdFst. Although...