DC FieldValueLanguage
dc.contributor.authorScheidl, Harald-
dc.contributor.authorFiel, Stefan-
dc.contributor.authorSablatnig, Robert-
dc.date.accessioned2020-06-27T18:27:20Z-
dc.date.issued2018-
dc.identifier.urihttps://resolver.obvsg.at/urn:nbn:at:at-ubtuw:3-3778-
dc.identifier.urihttp://hdl.handle.net/20.500.12708/924-
dc.descriptionThe final publication is available via <a href="https://doi.org/10.1109/ICFHR-2018.2018.00052" target="_blank">https://doi.org/10.1109/ICFHR-2018.2018.00052</a>.-
dc.description.abstractRecurrent Neural Networks (RNNs) are used for sequence recognition tasks such as Handwritten Text Recognition (HTR) or speech recognition. If trained with the Connectionist Temporal Classification (CTC) loss function, the output of such a RNN is a matrix containing character probabilities for each time-step. A CTC decoding algorithm maps these character probabilities to the final text. Token passing is such an algorithm and is able to constrain the recognized text to a sequence of dictionary words. However, the running time of token passing depends quadratically on the dictionary size and it is not able to decode arbitrary character strings like numbers. This paper proposes word beam search decoding, which is able to tackle these problems. It constrains words to those contained in a dictionary, allows arbitrary non-word character strings between words, optionally integrates a word-level language model and has a better running time than token passing. The proposed algorithm outperforms best path decoding, vanilla beam search decoding and token passing on the IAM and Bentham HTR datasets. An open-source implementation is provided.en
dc.description.sponsorshipEuropean Union's Horizon 2020-
dc.languageEnglish-
dc.language.isoen-
dc.rights.urihttp://rightsstatements.org/vocab/InC-EDU/1.0/ CC-
dc.subjectconnectionist temporal classificationen
dc.subjectdecodingen
dc.subjectlanguage modelen
dc.subjectrecurrent neural networken
dc.subjectspeech recognitionen
dc.subjecthandwritten text recognitionen
dc.titleWord Beam Search: A Connectionist Temporal Classification Decoding Algorithmen
dc.typeInproceedingsen
dc.typeKonferenzbeitragde
dc.relation.publication16th International Conference on Frontiers in Handwriting Recognition (ICFHR 2018)-
dc.relation.isbn9781538658758-
dc.relation.grantno674943-
dc.rights.holder2018 IEEE-
tuw.relation.publisherNiagara Falls-
tuw.versionam-
tuw.publication.orgunitE193 - Institut für Visual Computing and Human-Centered Technology-
tuw.publisher.doi10.1109/ICFHR-2018.2018.00052-
dc.identifier.libraryidAC15148695-
dc.identifier.urnurn:nbn:at:at-ubtuw:3-3778-
dc.rights.identifierIn Copyright - Educational Use Permitted-
item.openairetypeInproceedings-
item.openairetypeKonferenzbeitrag-
item.openaccessfulltextOpen Access-
item.languageiso639-1en-
item.openairecristypehttp://purl.org/coar/resource_type/c_18cf-
item.openairecristypehttp://purl.org/coar/resource_type/c_18cf-
item.grantfulltextopen-
item.fulltextwith Fulltext-
item.cerifentitytypePublications-
item.cerifentitytypePublications-
Appears in Collections:Conference Paper

Files in this item:


Page view(s)

903
checked on Jul 31, 2021

Download(s)

4,473
checked on Jul 31, 2021

Google ScholarTM

Check


Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.