Posts

• Low-energy dockerfile for ctc-forced-aligner

  I have to use docker.

  Oct 15, 2024

• Low-energy dockerfile for WhisperX

  I have to use docker.

  Oct 14, 2024

• Get data from Folkets Swedish-English

  Just pronunciation information

  Oct 12, 2024

• Trying to use pocketsphinx to word align

  Because timing accuracy in ASR is getting progressively worse, look backwards

  Oct 6, 2024

• Clean audio

  Zeroing out segments from a tsv

  Oct 3, 2024

• Espeak to/from CMUdict

  Imperfect mapping between imperfect phonesets

  Sep 24, 2024

• TextGrid to Label Studio

  Via the API

  Sep 20, 2024

• WhisperX JSON to Label Studio

  Via the API

  Sep 19, 2024

• Google ASR JSON to TextGrid

  For editing. Because it's needed.

  Sep 6, 2024

• WhisperX prompted

  Trying to get better output from WhisperX by prompting and using a wav2vec model trained on more spontaneous speech

  Aug 29, 2024

• Expand MFA lexicon

  MFA has phonological rules, but the implementation is useless. This approximates phonological rules for our speakers

  Aug 27, 2024

• Convert HSI data to fairseq

  Maybe a fine-tuned wav2vec model will work better with WhisperX

  Aug 22, 2024

• Textgrid durations

  Get max/min segment durations from textgrids

  Aug 19, 2024

• Mapping (CMUdict) ARPAbet to Oculus visemes

  For a colleague

  Aug 10, 2024

• MT trained on crawled data still sucks

  Common Crawl contains a lot of Google Translate output. See if you can guess the source material

  Aug 2, 2024