We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.AS

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Audio and Speech Processing

Title: Data augmentation using prosody and false starts to recognize non-native children's speech

Abstract: This paper describes AaltoASR's speech recognition system for the INTERSPEECH 2020 shared task on Automatic Speech Recognition (ASR) for non-native children's speech. The task is to recognize non-native speech from children of various age groups given a limited amount of speech. Moreover, the speech being spontaneous has false starts transcribed as partial words, which in the test transcriptions leads to unseen partial words. To cope with these two challenges, we investigate a data augmentation-based approach. Firstly, we apply the prosody-based data augmentation to supplement the audio data. Secondly, we simulate false starts by introducing partial-word noise in the language modeling corpora creating new words. Acoustic models trained on prosody-based augmented data outperform the models using the baseline recipe or the SpecAugment-based augmentation. The partial-word noise also helps to improve the baseline language model. Our ASR system, a combination of these schemes, is placed third in the evaluation period and achieves the word error rate of 18.71%. Post-evaluation period, we observe that increasing the amounts of prosody-based augmented data leads to better performance. Furthermore, removing low-confidence-score words from hypotheses can lead to further gains. These two improvements lower the ASR error rate to 17.99%.
Subjects: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL)
Cite as: arXiv:2008.12914 [eess.AS]
  (or arXiv:2008.12914v1 [eess.AS] for this version)

Submission history

From: Mittul Singh [view email]
[v1] Sat, 29 Aug 2020 05:32:32 GMT (450kb,D)

Link back to: arXiv, form interface, contact.