Lightly supervised learning from a damaged natural speech corpus

C. Fox, T. Hain

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Large corpora of transcribed speech are rare and expensive to acquire, but valuable for ASR systems. Of current research interest are corpora of natural speech, i.e. far-field recordings of multiple speakers in noisy environments. In the big data era there are many speech transcriptions collected for purposes other than ASR, which omit features required by typical ASR systems such as timing information. If we could recover training data from such 'found' corpora this would open up large new resources for ASR research. We present a case study for this type of data recovery - becoming known as 'lightly supervised learning' - for a highly damaged corpus called Family Life. We use a novel comparison of a parallel decode and forced audio alignment to iteratively select and grow good data. Family Life also has unusual data mislabelling problems which can be addressed by an integrated tfidf approach. These methods reduce WER on the corpus from 83.0 to 57.2. We also discuss a probabilistic loose string alignment approach which removes untranscribed 'icebreaker' speech.
Original languageEnglish
Title of host publicationAcoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on
PublisherInstitute of Electrical and Electronics Engineers
Pages8086-8090
Number of pages5
DOIs
Publication statusPublished - 2013

Keywords / Materials (for Non-textual outputs)

  • learning (artificial intelligence)
  • natural language processing
  • speech recognition
  • ASR systems
  • Family Life
  • WER
  • data mislabelling problem
  • data recovery
  • forced audio alignment
  • highly damaged corpus
  • integrated tfidf approach
  • lightly supervised learning
  • natural speech
  • parallel decode
  • probabilistic loose string alignment approach
  • speech transcriptions
  • training data
  • transcribed speech
  • Acoustics
  • Error analysis
  • Hidden Markov models
  • Interviews
  • Speech
  • Timing
  • Training

Fingerprint

Dive into the research topics of 'Lightly supervised learning from a damaged natural speech corpus'. Together they form a unique fingerprint.

Cite this