The goal of this project is to extend PocketSphinx so that it can calculate and provide alignments and acoustic scores for diphones similarly to how it already does for phonemes. This will help with learner analytics and adaptive instruction for pronunciation evaluation, assessment, and remediation, and can increase speech recognition accuracy and speech recognition system generality. Moreover, diphones may be helpful for new acoustic models, and can inform the construction of different kinds of language models.

Organization

Student

Pavel Denisov

Mentors

  • James Salsman
  • Christopher Bader
close

2017