The applicable script and parameters will then be specified in a file called cmd.sh located at the top level of your corpus’ training directory. Kaldi’s wrapper scripts are run.pl,, and, along with a few others we won’t discuss here. Kaldi provides a wrapper to implement this parallelization so that each of the computational steps can take advantage of the multiple processors. The number of jobs or splits in the dataset will be specified later in the training and alignment steps. Both training and alignment can be made more efficient by splitting the dataset into smaller chunks and processing them in parallel. Training can be computationally expensive however, if you have multiple processors/cores or even multiple machines, there are ways to speed it up significantly. You should notice fairly logical and linguistically motivated divisions among the phones. It is worth taking a look at this file to see how the model may be learning more about a phoneme’s contextual information. phones is a directory containing many additional files, including the extra_questions.txt file mentioned in section 5.3. The new files located in data/lang are L.fst, L_disambig.fst, oov.int, oov.txt, phones.txt, topo, words.txt, and phones. These must now point to different directories. Note that some older versions of Kaldi allowed the source and tmp directories to refer to the same location. Make sure that this entry and its corresponding phone ( oov) are entered in lexicon.txt and the phone is listed in silence_phones.txt. The second argument refers to lexical entry (word) for a “spoken noise” or “out of vocabulary” phone. # where the underlying argument structure is: Utils /prepare_lang.sh data /local /lang 'OOV' data /local / data /lang
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |