EUVIMEDEUVIMED entdecken
Crossref

Challenges in Natural Language Processing for Malayalam: A Low-Resource Language Perspective

A. G Saithya

International Journal of Multilingualism and Languages for Specific purposes · 31.12.2025 · 7 (02) · 104-123

DOI: 10.52919/ijmlsp.v7i02.123 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Multilingualism and Languages for Specific purposes, online erschienen am 31.12.2025. Verfasst von A. G Saithya. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Malayalam, a prominent Dravidian language chiefly spoken in the Indian state of Kerala, poses distinct difficulties for Natural Language Processing (NLP). Although it boasts over 35 million speakers, the language is still underrepresented in computational linguistics studies relative to high-resource languages like English or Hindi. This paper systematically examines the structural, sociolinguistic, and technological barriers to developing effective NLP systems for Malayalam, with particular attention to its agglutinative and highly inflectional morphology, diglossia and register variation, prevalent code-mixing with English, and non-standard Romanized writing practices known as Manglish. Through a theoretical and analytical approach, the study synthesizes and critically evaluates findings from Malayalam-specific and low-resource NLP research and assesses the applicability and limitations of computational methods such as multilingual pre-trained models, transliteration systems, and transfer learning techniques in the Malayalam context. The analysis finds that morphological complexity, limited annotated corpora, and inconsistent orthographic conventions significantly reduce model performance and limit the effectiveness of standard NLP pipelines. While multilingual and cross-lingual models partially address these issues by leveraging resource-rich languages, their effectiveness for Malayalam remains constrained by insufficient language-specific data and inadequate handling of morphological and register variation. This paper contributes by clearly identifying the core linguistic and resource-based challenges in Malayalam NLP and by proposing practical, research-informed strategies for corpus development, model adaptation, and hybrid linguistic-computational approaches. By situating Malayalam within broader discussions on low-resource and multilingual NLP, the study provides insights to inform future research and model development for Malayalam and other underrepresented languages, thereby promoting more inclusive and linguistically diverse digital technologies. The paper also highlights Malayalam's typological value as a morphologically rich Dravidian language, contending that knowledge from its analysis may guide the development of natural language processing (NLP) systems for structurally related languages. It also describes possible directions for future research, such as focused corpus growth, better annotation techniques, and low-resource evaluation systems. The report aims to close the gap between descriptive studies and workable solutions by outlining these directions.

Zitieren

APA

Saithya, A. G. (2025). Challenges in Natural Language Processing for Malayalam: A Low-Resource Language Perspective. International Journal of Multilingualism and Languages for Specific purposes, 7(02), 104-123. https://doi.org/10.52919/ijmlsp.v7i02.123

Vancouver

Saithya AG. Challenges in Natural Language Processing for Malayalam: A Low-Resource Language Perspective. International Journal of Multilingualism and Languages for Specific purposes. 2025;7(02):104-123. doi:10.52919/ijmlsp.v7i02.123

BibTeX
@article{Saithya2025,
  author = {Saithya, A. G},
  title = {Challenges in Natural Language Processing for Malayalam: A Low-Resource Language Perspective},
  journal = {International Journal of Multilingualism and Languages for Specific purposes},
  year = {2025},
  volume = {7},
  number = {02},
  pages = {104--123},
  doi = {10.52919/ijmlsp.v7i02.123}
}
Kostenlos starten →