Part of Speech Tagging for Setswana African Language

Dibitso, Mary Ambrossine; Owolawi, Pius A.; Ojo, Sunday O.

doi:10.1109/imitec45504.2019.9015871

Cited by 2 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite a long history of work on individual NLP tasks on African languages (Adedjouma Sèmiyou et al, 2012;Dibitso et al, 2019;Schlunz et al, 2016;Pauw et al, 2006;Onyenwe et al, 2014;Hunegnaw et al, 2021;Orimaye et al, 2012;Eiselen, 2016;, the lack of freely available and aggregated models made it difficult for languages to build off of each other.…”

Section: Related Workmentioning

confidence: 99%

Mini But Mighty: Efficient Multilingual Pretraining with Linguistically-Informed Data Selection

Ogunremi,

Jurafsky,

Manning

2023

Findings of the Association for Computational Linguistics: EACL 2023

View full text Add to dashboard Cite

With the prominence of large pretrained language models, low-resource languages are rarely modelled monolingually and become victims of the "curse of multilinguality" in massively multilingual models. Recently, Afri-BERTa showed that training transformer models from scratch on 1GB of data from many unrelated African languages outperforms massively multilingual models on downstream NLP tasks. Here we extend this direction, focusing on the use of related languages. We propose that training on smaller amounts of data but from related languages could match the performance of models trained on large, unrelated data. We test our hypothesis on the Niger-Congo family and its Bantu and Volta-Niger sub-families, pretraining models with data solely from Niger-Congo languages and finetuning on 4 downstream tasks: NER, partof-speech tagging, sentiment analysis and text classification. We find that models trained on genetically related languages achieve equal performance on downstream tasks in low-resource languages despite using less training data. We recommend selecting training data based on language-relatedness when pretraining language models for low-resource languages.

show abstract

Section: Related Workmentioning

confidence: 99%

Mini But Mighty: Efficient Multilingual Pretraining with Linguistically-Informed Data Selection

Ogunremi,

Jurafsky,

Manning

2023

Findings of the Association for Computational Linguistics: EACL 2023

View full text Add to dashboard Cite

show abstract

SPECIL: Spell Error Corpus for the Indonesian Language

Yanfi,

Setiawan,

Soeparno

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In this study, we present the first spell error corpus for the Indonesian Language (SPECIL). This corpus provides a comprehensive resource for researchers and practitioners to detect and correct spelling errors in Bahasa Indonesia (Indonesian). It should be emphasized that currently, there is no recognized corpus for identifying spelling mistakes in the Indonesian language that has been officially released or made accessible. This study also provides a systematic literature review to identify resources and methodologies for building a corpus for spelling error detection and correction in Indonesia. A corpus was created using a combination of manual and automatic methods. The results of this study are a review of publications relating to corpora and spelling, the novel algorithm of six types of spelling errors, and the production of a corpus comprising over 180,000 tokens in 21,500 sentences, including non-word, real-word, and punctuation errors. Using the developed corpus, various Natural Language Processing (NLP) models, including spell checkers and language models, can be trained and tested to enhance their accuracy and effectiveness in identifying and rectifying errors in Indonesian texts. Moreover, the corpus can be used to develop and evaluate new algorithms and techniques for spelling error detection and correction in Indonesia. The SPECIL corpus is publicly available and accessible. It is expected that SPECIL will inspire further research in this area and facilitate the development of more accurate and effective spelling error detection and correction tools in Indonesian language.

show abstract

Part of Speech Tagging for Setswana African Language

Cited by 2 publications

References 13 publications

Mini But Mighty: Efficient Multilingual Pretraining with Linguistically-Informed Data Selection

Mini But Mighty: Efficient Multilingual Pretraining with Linguistically-Informed Data Selection

SPECIL: Spell Error Corpus for the Indonesian Language

Contact Info

Product

Resources

About