Using Part of Speech N-Grams for Improving Automatic Speech Recognition of Polish

Pohl, Aleksander; Ziółko, Bartosz

doi:10.1007/978-3-642-39712-7_38

Cited by 8 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Subsequent words are predicted based on previous words, and the meaning of the utterance is discovered based on the local context [17]. Statistical approaches are appropriate for English, which is positional and has a specific sentence formation (most commonly subject-verb-object (SVO) [18]).…”

Section: A Conventional Asrmentioning

confidence: 99%

End-To-End deep neural models for Automatic Speech Recognition for Polish Language

Pondel-Sycz,

Pietrzak,

Szymla

2024

International Journal of Electronics and Telecommunications

View full text Add to dashboard Cite

This article concerns research on deep learning models (DNN) used for automatic speech recognition (ASR). In such systems, recognition is based on Mel Frequency Cepstral Coefficients (MFCC) acoustic features and spectrograms. The latest ASR technologies are based on convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers. The article presents an analysis of modern artificial intelligence algorithms adapted for automatic recognition of the Polish language. The differences between conventional architectures and ASR DNN End-To-End (E2E) models are discussed. Preliminary tests of five selected models (QuartzNet, FastConformer, Wav2Vec 2.0 XLSR, Whisper and ESPnet Model Zoo) on Mozilla Common Voice, Multilingual LibriSpeech and VoxPopuli databases are demonstrated. Tests were conducted for clean audio signal, signal with bandwidth limitation and degraded. The tested models were evaluated on the basis of Word Error Rate (WER).

show abstract

Section: A Conventional Asrmentioning

confidence: 99%

End-To-End deep neural models for Automatic Speech Recognition for Polish Language

Pondel-Sycz,

Pietrzak,

Szymla

2024

International Journal of Electronics and Telecommunications

View full text Add to dashboard Cite

show abstract

“…However, they are mostly several years old and based on nonneural network methods. Some examples include systems based on hidden Markov models [30], k-nearest neighbours [31], and speech n-grams [32]. Results achieved with these methods are not satisfactory; therefore, they were not considered further in our project.…”

Section: Speech Recognition In Low-resource Settingsmentioning

confidence: 99%

Generalisation Gap of Keyword Spotters in a Cross-Speaker Low-Resource Scenario

Lepak

Radzikowski

Nowak

et al. 2021

Sensors

View full text Add to dashboard Cite

Models for keyword spotting in continuous recordings can significantly improve the experience of navigating vast libraries of audio recordings. In this paper, we describe the development of such a keyword spotting system detecting regions of interest in Polish call centre conversations. Unfortunately, in spite of recent advancements in automatic speech recognition systems, human-level transcription accuracy reported on English benchmarks does not reflect the performance achievable in low-resource languages, such as Polish. Therefore, in this work, we shift our focus from complete speech-to-text conversion to acoustic similarity matching in the hope of reducing the demand for data annotation. As our primary approach, we evaluate Siamese and prototypical neural networks trained on several datasets of English and Polish recordings. While we obtain usable results in English, our models’ performance remains unsatisfactory when applied to Polish speech, both after mono- and cross-lingual training. This performance gap shows that generalisation with limited training resources is a significant obstacle for actual deployments in low-resource languages. As a potential countermeasure, we implement a detector using audio embeddings generated with a generic pre-trained model provided by Google. It has a much more favourable profile when applied in a cross-lingual setup to detect Polish audio patterns. Nevertheless, despite these promising results, its performance on out-of-distribution data are still far from stellar. It would indicate that, in spite of the richness of internal representations created by more generic models, such speech embeddings are not entirely malleable to cross-language transfer.

show abstract

“…This were recently replaced with deep neural network for popular languages like English. However most of the literature devoted to the Polish language considers n-gram models [12,18,19,21]. Brocki et al [4] showed that a simple neural network (5 context words with 50 dimensional embeddings and one hidden layer) greatly outperforms a 4-gram solution on a Polish corpus.…”

Section: Related Workmentioning

confidence: 99%

Universal Language Model Fine-Tuning with Subword Tokenization for Polish

Czapla,

Howard,

Kardas

2018

Preprint

View full text Add to dashboard Cite

Universal Language Model for Fine-tuning [6] (ULMFiT) is one of the first NLP methods for efficient inductive transfer learning. Unsupervised pretraining results in improvements on many NLP tasks for English. In this paper, we describe a new method that uses subword tokenization to adapt ULMFiT to languages with high inflection. Our approach results in a new state-of-the-art for the Polish language, taking first place in Task 3 of PolEval'18. After further training, our final model outperformed the second best model by 35%. We have open-sourced our pretrained models and code. 3

show abstract

Using Part of Speech N-Grams for Improving Automatic Speech Recognition of Polish

Cited by 8 publications

References 10 publications

End-To-End deep neural models for Automatic Speech Recognition for Polish Language

End-To-End deep neural models for Automatic Speech Recognition for Polish Language

Generalisation Gap of Keyword Spotters in a Cross-Speaker Low-Resource Scenario

Universal Language Model Fine-Tuning with Subword Tokenization for Polish

Contact Info

Product

Resources

About