Exploring Emotion Detection and Sentiment Analysis of Texts in Low-Resource Languages: Techniques Challenges, Emerging Trends and Future Directions

Abstract

Emotion detection and sentiment analysis in low-resource languages have 
gained significant attention in natural language processing (NLP). This is due to 
their importance in understanding user-generated content, especially on social 
media platforms. However, these tasks face numerous challenges due to the 
scarcity of labelled data, complex linguistic structures, and limited computational 
resources. This paper explores the state-of-the-art methodologies for emotion 
detection and sentiment analysis in low-resource languages. We identify key 
open problems such as poor analytical accuracy due to limited datasets, the 
complicated morphology of native languages, and the lack of suitable linguistic 
resources. Additionally, we discuss the use of machine learning and deep 
learning models, including convolutional neural networks (CNNs), recurrent 
neural networks (RNNs), deep belief networks (DBNs), and lexicon-based 
approaches, as potential solutions to these challenges. These approaches 
leverage techniques such as transfer learning, pre-trained embeddings, and 
hybrid models to mitigate the limitations of small datasets. Finally, to further 
address these problems, we present emerging trends in research in this field, a 
combination of advanced models and data augmentation techniques to 
enhance the accuracy and robustness of emotion detection and sentiment 
analysis in these languages.