Dynamic out-of-vocabulary word registration to language model for speech recognition

Kitaoka, Norihide; Chen, Bohan; Obashi, Yuya

doi:10.1186/s13636-020-00193-1

Cited by 5 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This, in turn, results in the hypothesis always containing words in the lexicon leading to errors in recognition. There have been efforts to detect OOVs using filler models [41,42,43] where there are placeholders for OOVs which can then be replaced with the extended vocabulary for improved recognition. Confidence measures are another indication of OOV presence.…”

Section: Oov Detection and Recoverymentioning

confidence: 99%

Low Resource Kannada Speech Recognition using Lattice Rescoring and Speech Synthesis

Murthy¹,

Sitaram²

2023

IJST

View full text Add to dashboard Cite

Objectives: Improving the accuracy of low resource speech recognition in a model trained on only 4 hours of transcribed continuous speech in Kannada language, using data augmentation. Methods: Baseline language model is augmented with unigram counts of words, that are present in the Wikipedia text corpus but absent in the baseline, for initial decoding. Lattice rescoring is then applied using the language model augmented with Wikipedia text. Speech synthesis-based augmentation with multi-speaker syllable-based synthesis, using voices in Kannada and cross-lingual Telugu languages, is employed. We synthesize basic syllables, syllables with consonant conjuncts, and words that contain syllables that are absent in the training speech, for Kannada language. Findings: An overall word error rate (WER) of 9.04% is achieved over a baseline WER of 40.93%. Language model augmentation and lattice rescoring gives an absolute improvement of 16.68%. Applying our method of syllable-based speech synthesis over language model augmentation and rescoring yields a total reduction of 31.89% in WER. The proposed approach of language model augmentation is memory efficient and consumes only 1/8 th the memory required for decoding with Wikipedia augmented language model (2 gigabytes versus 18 gigabytes) while giving comparable WER (22.95% for Wikipedia versus 24.25% for our method). Augmentation with synthesized syllables enhances the ability of the speech recognition model to recognize basic sounds thus improving recognition of out-of-vocabulary words to 90% and in-vocabulary words to 97%. Novelty: We propose novel methods of language model augmentation and synthesis-based augmentation to achieve low WER for a speech recognition model trained on only 4 hours of continuous speech. Obtaining high recognition accuracy (or low WER) for very small speech corpus is a challenge. In this paper, we demonstrate that high accuracy can be achieved using data augmentation for a small corpus-based speech recognition.

show abstract

Section: Oov Detection and Recoverymentioning

confidence: 99%

Low Resource Kannada Speech Recognition using Lattice Rescoring and Speech Synthesis

Murthy¹,

Sitaram²

2023

IJST

View full text Add to dashboard Cite

show abstract

“…The ratio of the number of words in the corpus to the number of vocabulary is important, and in the case of "Decision to Leave", the ratio is about 4.12. To explain the ratio, we can use the analogy of Lego blocks: the fewer the types of blocks used (vocabulary) and the larger the structure created by stacking them (corpus), the easier it is to learn the order of application of the blocks [27]. Naturally, if a language model is created using only one movie script, the sentence generation performance is low, but it is reasonable for the purpose of obtaining word vectors and utilizing vector operations as in this study [28].…”

Section: Language Model -Structurementioning

confidence: 99%

Word Vectors for Criticism of a Korean Film - 'Decision to Leave' by Chanwook Park

Ko,

Paik

2023

Preprint

View full text Add to dashboard Cite

This research presents a novel approach to film analysis, leveraging word vectors to objectively evaluate movie critiques. Focusing on Director Chanwook Park’s award-winning film, ’Decision to Leave’, the study employs word vectors derived from the movie’s script of Korean text. Traditional critiques often emphasize contrasting elements and themes, but their subjective nature poses challenges in objective validation. To address this, we trained a language model using LSTM on the film’s script, obtaining word vectors that capture the essence of the narrative. These vectors were then used to perform various text analyses, including similarity and analogy operation for the keywords suggested by the critiques. By comparing the semantic relationships in the critiques with those derived from the word vectors, we could objectively validate the critiques’ assertions. Furthermore, we visualized the word vectors in a two-dimensional space, confirming the spatial relationships of key terms highlighted in critiques. The study underscores the potential of word vectors in providing a more objective lens for film analysis, bridging the gap between traditional film criticism and data-driven insights.

show abstract

A comprehensive survey on automatic speech recognition using neural networks

Dhanjal,

Singh

2023

Multimed Tools Appl

View full text Add to dashboard Cite

Dynamic out-of-vocabulary word registration to language model for speech recognition

Cited by 5 publications

References 16 publications

Low Resource Kannada Speech Recognition using Lattice Rescoring and Speech Synthesis

Low Resource Kannada Speech Recognition using Lattice Rescoring and Speech Synthesis

Word Vectors for Criticism of a Korean Film - 'Decision to Leave' by Chanwook Park

A comprehensive survey on automatic speech recognition using neural networks

Contact Info

Product

Resources

About