Filters








61 Hits in 7.8 sec

Towards One Model to Rule All: Multilingual Strategy for Dialectal Code-Switching Arabic ASR [article]

Shammur Absar Chowdhury, Amir Hussein, Ahmed Abdelali, Ahmed Ali
2021 arXiv   pre-print
Our findings demonstrate the strength of such a model by outperforming state-of-the-art monolingual dialectal Arabic and code-switching Arabic ASR.  ...  With the advent of globalization, there is an increasing demand for multilingual automatic speech recognition (ASR), handling language and dialectal variation of spoken content.  ...  Conclusions In this paper, we presented the first comprehensive study comparing multilingual ASR strategy to develop an E2E Arabic dialectal and code-switching ASR.  ... 
arXiv:2105.14779v2 fatcat:k5ovhezrmjhs7megwwi7gelxve

Towards One Model to Rule All: Multilingual Strategy for Dialectal Code-Switching Arabic ASR

Shammur Absar Chowdhury, Amir Hussein, Ahmed Abdelali, Ahmed Ali
2021 Interspeech 2021   unpublished
This is the first study to benchmark the performance of a multilingual ASR for dialectal and code-switching Arabic test sets. The proposed model outperforms current Arabic state-of-the-art E2E  ...  Our findings demonstrate the strength of such a model by outperforming state-of-the-art monolingual dialectal Arabic and code-switching Arabic ASR.  ...  Conclusions In this paper, we presented the first comprehensive study comparing multilingual ASR strategy to develop an E2E Arabic dialectal and code-switching ASR.  ... 
doi:10.21437/interspeech.2021-1809 fatcat:u34wfklwmjb4fl4va5bn6gh34y

A Survey of Code-switched Speech and Language Processing [article]

Sunayana Sitaram, Khyathi Raghavi Chandu, Sai Krishna Rallabandi, Alan W Black
2020 arXiv   pre-print
We motivate why processing code-switched text and speech is essential for building intelligent agents and systems that interact with users in multilingual communities.  ...  This survey reviews computational approaches for code-switched Speech and Natural Language Processing.  ...  Arabic Dialects are often mixed with Modern Standard Arabic. The distinction between languages and dialects is hard to define, but we see that code-switching appears with dialects too.  ... 
arXiv:1904.00784v3 fatcat:r5tsg4kdnfbtnndae523c32pta

Review of distinctive phonetic features and the Arabic share in related modern research

Yousef ALOTAIBI, Ali MEFTAH
2013 Turkish Journal of Electrical Engineering and Computer Sciences  
Based on the literature search conducted for this paper, we conclude that although the Arabic language is a very important and old Semitic language, hitherto it has suffered from a lack of modern research  ...  In addition, we highlight the background of Arabic language phonology compared to 2 well-known languages to enhance the current knowledge about this narrow language discipline.  ...  [33] worked on phone set construction based on context-sensitive articulatory attributes for code-switching speech recognition.  ... 
doi:10.3906/elk-1112-29 fatcat:wu2yuv2janemdjtnp6ohec3vai

Multilingual spoken language processing

P. Fung, T. Schultz
2008 IEEE Signal Processing Magazine  
Fung and Cheung [7] suggest predicting the code-switch location either from empirical data or from linguistic rules.  ...  In addition, multilingual acoustic models allow for the construction of truly multilingual systems, which can handle code switching and cross-lingual pronunciation effects.  ... 
doi:10.1109/msp.2008.918417 fatcat:ezye4rngebdpphtis3szqdhvce

Data Augmentation for Speech Recognition in Maltese: A Low-Resource Perspective [article]

Carlos Mena and Andrea DeMarco and Claudia Borg and Lonneke van der Plas and Albert Gatt
2021 arXiv   pre-print
Our results show that combining the three data augmentation techniques studied here lead us to an absolute WER improvement of 15% without the use of a language model.  ...  Developing speech technologies is a challenge for low-resource languages for which both annotated and raw speech data is sparse. Maltese is one such language.  ...  The audiobooks portion of our pretraining data was generously provided by Merlin Publishers Ltd; special thanks are due to Chris Gruppetta for facilitating the acquisition process.  ... 
arXiv:2111.07793v1 fatcat:mykoon2fdbdsbn3xl3j66enenq

Systematic Literature Review of Dialectal Arabic: Identification and Detection

Ashraf Elnagar, Sane Yagi, Ali Bou Nassif, Ismail Shahin, Said A. Salloum
2021 IEEE Access  
Proper instructions on the detection of Arabic code switching in regard to tokens and words have been suggested in [68] .  ...  Alsalka, "Compression versus traditional machine learning classifiers to detect code-switching in varieties and dialects: Arabic as a case study," Nat. Lang. Eng., pp. 1-14.  ...  Content may change prior to final publication.  ... 
doi:10.1109/access.2021.3059504 fatcat:d7dkxmdehzcq5d7fej7icyy6rq

Mixed Script Identification Using Automated DNN Hyperparameter Optimization

Muhammad Yasir, Li Chen, Amna Khatoon, Muhammad Amir Malik, Fazeel Abid, Ahmed Mostafa Khalil
2021 Computational Intelligence and Neuroscience  
Moreover, this study addresses the issues related to multilingual environments, such as Roman words merged with English characters, generative spellings, and phonetic typing.  ...  This study tackles the challenge of mixed script identification for mixed-code dataset consisting of Roman Urdu, Hindi, Saraiki, Bengali, and English.  ...  Arabic dialect in code-mixed tweets,” in Proceedings of the [5] J. Chung, C. Gulcehre, K. Cho, and Y.  ... 
doi:10.1155/2021/8415333 pmid:34925496 pmcid:PMC8683192 fatcat:6ilsi3lsmjclzjjbgby7wuioui

Multimodal Machine Translation through Visuals and Speech

Umut Sulubacak, Ozan Caglayan, Stig-Arne Grönroos, Aku Rouhe, Desmond Elliott, Lucia Specia, Jörg Tiedemann
2019 Zenodo  
The paper concludes with a discussion of directions for future research in these areas: the need for more expansive and challenging datasets, for targeted evaluations of model performance, and for multimodality  ...  These tasks are distinguished from their monolingual counterparts of speech recognition, image captioning, and video captioning by the requirement of models to generate outputs in a different language.  ...  We would also like to thank Maarit Koponen for her valuable feedback and her help in establishing our discussions of machine translation evaluation.  ... 
doi:10.5281/zenodo.3690791 fatcat:otdy5i33fzfsnnbb3xgb6zph6q

Multimodal machine translation through visuals and speech

Umut Sulubacak, Ozan Caglayan, Stig-Arne Grönroos, Aku Rouhe, Desmond Elliott, Lucia Specia, Jörg Tiedemann
2020 Machine Translation  
The paper concludes with a discussion of directions for future research in these areas: the need for more expansive and challenging datasets, for targeted evaluations of model performance, and for multimodality  ...  These tasks are distinguished from their monolingual counterparts of speech recognition, image captioning, and video captioning by the requirement of models to generate outputs in a different language.  ...  We would also like to thank Maarit Koponen for her valuable feedback and her help in establishing our discussions of machine translation evaluation.  ... 
doi:10.1007/s10590-020-09250-0 fatcat:jod3ghcsnnbipotcqp6sme4lna

Searching spontaneous conversational speech

Franciska de Jong, Douglas W. Oard, Roeland Ordelman, Stephan Raaijmakers
2007 SIGIR Forum  
over a reasonable range of transcription accuracy. • Sufficiently accurate Large-Vocabulary Continuous Speech Recognition (LVCSR) systems could be built for the planned speech of news announcers.  ...  Three factors were key to this success: • Broadcast news has a "story" structure that resembles written documents. • The redundancy present in human language meant that search effectiveness held up well  ...  I thank Qian and John as well as Inderjeet Mani and Lisa Ferro for their comments on drafts of this paper. Acknowledgements The author is grateful to Douglas W.  ... 
doi:10.1145/1328964.1328982 fatcat:wwpzqq7ndrfedh4imhoznvccl4

D2.3 Software and demonstration of human-like content description generation

Doukhan, Guo, Harrando, Kurimo, Laaksonen, Lindgren, Lindh-Knuutila, Lisena, Pehlivan Tort, Reboud, Rouhe, Troncy (+1 others)
2020 Zenodo  
The ultimate goal of the development is to be able to generate human-like content descriptions, in which the persons, objects, actions and environments, both seen and heard, are described in a sophisticated  ...  This deliverable describes the last development iteration of the joint collection of libraries and tools for multimodal content analysis and description from AALTO, EURECOM, INA, Lingsoft, LLS and Limecraft  ...  We noticed that language embeddings on the Arabic dialect dataset MGB-3 are easier to classify with GNB when we allow an open task approach where language embedding model is trained on all available, three  ... 
doi:10.5281/zenodo.4964391 fatcat:ertkzz2wbjajjlavw4iljlbmaq

Getting Past the Language Gap: Innovations in Machine Translation [chapter]

Rodolfo Delmonte
2012 Mobile Speech and Advanced Natural Language Solutions  
Its aim is the development of open-source MT technologies applicable to all language pairs within Europe, based on hybrid designs combining statistical and rule-based methods.  ...  On the one hand, MERT is not designed for models with rich features and therefore leads to translations of unstable quality in such scenarios.  ...  For instance, Popovicá nd Ney (2006) propose combining that strategy with the use of hand-crafted rules to reorder the SL sentences to match the structure of the TL.  ... 
doi:10.1007/978-1-4614-6018-3_6 fatcat:2njkc6meabhaxosl4wircumfjm

Probabilistic Modelling of Morphologically Rich Languages [article]

Jan A. Botha
2015 arXiv   pre-print
In the context of n-gram language modelling, we formulate a new Bayesian model that relies on the decomposition of compound words to attain better smoothing, and we develop a new distributed language model  ...  In both cases, we show that accounting for word sub-structure improves the models' intrinsic performance and provides benefits when applied to other tasks, including machine translation.  ...  As an example of FLMs using non-morphological information, Adel et al. (2013) applied them successfully to the problem of code-switching by incorporating a language-ID factor for each word.6 Another  ... 
arXiv:1508.04271v1 fatcat:6qhsfdbvt5emfiaumtwh2pzs7m

Automatic Speech Recognition for Low-resource Languages and Accents Using Multilingual and Crosslingual Information

Ngoc Thang Vu
2014
Under application aspects, this thesis also includes research work on non-native and Code-Switching speech.  ...  We focus on finding approaches which allow using data from multiple languages to improve the performance for those languages on different levels, such as feature extraction, acoustic modeling and language  ...  Multilingual ASR for Code-Switching speech In this thesis, we built an ASR system for conversational Mandarin-English Code-Switching speech.  ... 
doi:10.5445/ir/1000041124 fatcat:7kmt7i7tlnglfpdou6346spcky
« Previous Showing results 1 — 15 out of 61 results