Neural spelling correction: translating incorrect sentences to correct sentences for multimedia

Park, Chanjun; Kim, Kuekyeng; Yang, Yeongwook; Kang, Minho; Lim, Heuiseok

doi:10.1007/s11042-020-09148-2

Cited by 16 publications

(24 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, the BLEU and GLEU scores of the Filter model were higher than those of the existing spelling correction model proposed by Park et al (2020a) by 5.94 and 15.15, respectively. These results show that our post-processor can achieve higher performance in spelling correction.…”

Section: Quantitative Resultsmentioning

confidence: 67%

“…Base refers to the BLEU score between the source and target sentences; we leveraged it as the baseline for assessing the performance improvement. In addition, we compared the performance with that of the Korean spelling error correction model proposed by Park et al (2020a), who performed ASR post-processing experiments and published the model as a demo system † . This study focused on Korean spelling error correction that is not specialized in ASR post-processing.…”

Section: Quantitative Resultsmentioning

confidence: 99%

“…The fourth limitation is related to spelling errors, which frequently occur in the ASR result. Although previous studies (Kiyono et al, 2019;Choe et al, 2019;Park et al, 2020a) have investigated spelling correction, few have focused on ASR.…”

Section: Spelling Errorsmentioning

confidence: 99%

See 2 more Smart Citations

BTS: Back TranScription for Speech-to-Text Post-Processor using Text-to-Speech-to-Text

Park¹,

Seo²,

Lee³

et al. 2021

Proceedings of the 8th Workshop on Asian Translation (WAT2021)

Self Cite

View full text Add to dashboard Cite

With the growing popularity of smart speakers, such as Amazon Alexa, speech is becoming one of the most important modes of humancomputer interaction. Automatic speech recognition (ASR) is arguably the most critical component of such systems, as errors in speech recognition propagate to the downstream components and drastically degrade the user experience. A simple and effective way to improve the speech recognition accuracy is to apply automatic post-processor to the recognition result. However, training a post-processor requires parallel corpora created by human annotators, which are expensive and not scalable. To alleviate this problem, we propose Back TranScription (BTS), a denoising-based method that can create such corpora without human labor. Using a raw corpus, BTS corrupts the text using Text-to-Speech (TTS) and Speech-to-Text (STT) systems. Then, a postprocessing model can be trained to reconstruct the original text given the corrupted input. Quantitative and qualitative evaluations show that a post-processor trained using our approach is highly effective in fixing non-trivial speech recognition errors such as mishandling foreign words. We present the generated parallel corpus and post-processing platform to make our results publicly available.

show abstract

Section: Quantitative Resultsmentioning

confidence: 67%

Section: Quantitative Resultsmentioning

confidence: 99%

See 1 more Smart Citation

BTS: Back TranScription for Speech-to-Text Post-Processor using Text-to-Speech-to-Text

Park¹,

Seo²,

Lee³

et al. 2021

Proceedings of the 8th Workshop on Asian Translation (WAT2021)

Self Cite

View full text Add to dashboard Cite

show abstract

“…The S2S methodology corrects errors in the same way as the machine translation process Baskar et al, 2019;Park et al, 2020a). Based on the S2S model, the STT result is vectorized using an encoder and the vector is then decoded to generate a human-modified STT sentence.…”

Section: Sequence-to-sequence (S2s) Methodologymentioning

confidence: 99%

Proceedings of the 8th Workshop on Asian Translation (WAT2021)

Nakazawa¹,

Goto²

2021

View full text Add to dashboard Cite

Many Asian countries are rapidly growing these days and the importance of communicating and exchanging the information with these countries has intensified. To satisfy the demand for communication among these countries, machine translation technology is essential.Machine translation technology has rapidly evolved recently and it is seeing practical use especially between European languages. However, the translation quality of Asian languages is not that high compared to that of European languages, and machine translation technology for these languages has not reached a stage of proliferation yet. This is not only due to the lack of the language resources for Asian languages but also due to the lack of techniques to correctly transfer the meaning of sentences from/to Asian languages. Consequently, a place for gathering and sharing the resources and knowledge about Asian language translation is necessary to enhance machine translation research for Asian languages.

show abstract

“…Therefore, this paper aims to develop an algorithm for name matching, that consider an approximate string-matching algorithm to allow dealing with possible technical or computational errors. Such matching algorithms have been used in several applications such as spelling correction (Park et al, 2020), linking databases (Hand & Christen, 2018), text retrieval (Abdulhayoglu, Thijs & Jeuris, 2016), handwriting recognition (Chowdhury, Bhattacharya & Parui, 2013), computational biology ''DNA'' (Berger, Waterman & Yu, 2020), and name recognition (Delgado et al, 2016), etc. Consequently, in this work, a new softened distance measure is proposed, based on the BI-DIST distance to increase the efficiency and accuracy of the name-matching method.…”

Section: Introductionmentioning

confidence: 99%

Soft Bigram distance for names matching

Hadwan

Al-Hagery

Al-Sanabani

et al. 2021

PeerJ Computer Science

View full text Add to dashboard Cite

Background Bi-gram distance (BI-DIST) is a recent approach to measure the distance between two strings that have an important role in a wide range of applications in various areas. The importance of BI-DIST is due to its representational and computational efficiency, which has led to extensive research to further enhance its efficiency. However, developing an algorithm that can measure the distance of strings accurately and efficiently has posed a major challenge to many developers. Consequently, this research aims to design an algorithm that can match the names accurately. BI-DIST distance is considered the best orthographic measure for names identification; nevertheless, it lacks a distance scale between the name bigrams. Methods In this research, the Soft Bigram Distance (Soft-Bidist) measure is proposed. It is an extension of BI-DIST by softening the scale of comparison among the name Bigrams for improving the name matching. Different datasets are used to demonstrate the efficiency of the proposed method. Results The results show that Soft-Bidist outperforms the compared algorithms using different name matching datasets.

show abstract

Neural spelling correction: translating incorrect sentences to correct sentences for multimedia

Cited by 16 publications

References 23 publications

BTS: Back TranScription for Speech-to-Text Post-Processor using Text-to-Speech-to-Text

BTS: Back TranScription for Speech-to-Text Post-Processor using Text-to-Speech-to-Text

Proceedings of the 8th Workshop on Asian Translation (WAT2021)

Soft Bigram distance for names matching

Contact Info

Product

Resources

About