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: Audio-attention discriminative language model for ASR rescoring

Abstract: End-to-end approaches for automatic speech recognition (ASR) benefit from directly modeling the probability of the word sequence given the input audio stream in a single neural network. However, compared to conventional ASR systems, these models typically require more data to achieve comparable results. Well-known model adaptation techniques, to account for domain and style adaptation, are not easily applicable to end-to-end systems. Conventional HMM-based systems, on the other hand, have been optimized for various production environments and use cases. In this work, we propose to combine the benefits of end-to-end approaches with a conventional system using an attention-based discriminative language model that learns to rescore the output of a first-pass ASR system. We show that learning to rescore a list of potential ASR outputs is much simpler than learning to generate the hypothesis. The proposed model results in 8% improvement in word error rate even when the amount of training data is a fraction of data used for training the first-pass system.
Comments: 4 pages, 1 figure, Accepted at ICASSP 2020
Subjects: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Machine Learning (cs.LG); Sound (cs.SD)
Cite as: arXiv:1912.03363 [eess.AS]
  (or arXiv:1912.03363v2 [eess.AS] for this version)

Submission history

From: Ankur Gandhe [view email]
[v1] Fri, 6 Dec 2019 22:09:07 GMT (121kb,D)
[v2] Tue, 18 Feb 2020 18:03:03 GMT (234kb,D)

Link back to: arXiv, form interface, contact.