Show simple item record

dc.contributor.authorRagni, A
dc.contributor.authorLi, Q
dc.contributor.authorGales, MJF
dc.contributor.authorWang, Y
dc.date.accessioned2019-01-12T00:32:10Z
dc.date.available2019-01-12T00:32:10Z
dc.date.issued2018
dc.identifier.isbn9781538643341
dc.identifier.issn2639-5479
dc.identifier.urihttps://www.repository.cam.ac.uk/handle/1810/287923
dc.description.abstractThe standard approach to assess reliability of automatic speech transcriptions is through the use of confidence scores. If accurate, these scores provide a flexible mechanism to flag transcription errors for upstream and downstream applications. One challenging type of errors that recognisers make are deletions. These errors are not accounted for by the standard confidence estimation schemes and are hard to rectify in the upstream and downstream processing. High deletion rates are prominent in limited resource and highly mismatched training/testing conditions studied under IARPA Babel and Material programs. This paper looks at the use of bidirectional recurrent neural networks to yield confidence estimates in predicted as well as deleted words. Several simple schemes are examined for combination. To assess usefulness of this approach, the combined confidence score is examined for untranscribed data selection that favours transcriptions with lower deletion errors. Experiments are conducted using IARPA Babel/Material program languages.
dc.description.sponsorshipALTA Institute, Cambridge University; The Office of the Director of National Intelligence (ODNI), Intelligence Advanced Research Projects Activity (IARPA) via Air Force Research Laboratory (AFRL)
dc.publisherIEEE
dc.titleConfidence Estimation and Deletion Prediction Using Bidirectional Recurrent Neural Networks
dc.typeConference Object
prism.endingPage211
prism.publicationDate2019
prism.publicationName2018 IEEE Spoken Language Technology Workshop, SLT 2018 - Proceedings
prism.startingPage204
dc.identifier.doi10.17863/CAM.35236
dcterms.dateAccepted2018-09-03
rioxxterms.versionofrecord10.1109/SLT.2018.8639678
rioxxterms.versionAM
rioxxterms.licenseref.urihttp://www.rioxx.net/licenses/all-rights-reserved
rioxxterms.licenseref.startdate2019-02-11
dc.contributor.orcidGales, Mark [0000-0002-5311-8219]
rioxxterms.typeConference Paper/Proceeding/Abstract
pubs.conference-name2018 IEEE Spoken Language Technology Workshop (SLT)
pubs.conference-start-date2018-12-18
cam.orpheus.successThu Nov 05 11:53:19 GMT 2020 - Embargo updated
pubs.conference-finish-date2018-12-21
rioxxterms.freetoread.startdate2020-02-11


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record