Human-annotated dataset for social media sentiment analysis for Albanian language

Area of Science:

• Social Media Analysis
• Natural Language Processing
• Public Health Communication

Background:

• Social media platforms are crucial for public opinion expression during crises like the COVID-19 pandemic.
• Limited labeled datasets exist for low-resource languages, hindering research.
• Public health communication strategies require understanding public sentiment.

Purpose of the Study:

• To introduce a novel, human-annotated dataset of social media comments related to COVID-19 in Kosovo.
• To support research in machine learning and affective computing for low-resource languages.
• To aid public agencies in understanding public health discourse.

Main Methods:

• Collected 10,132 comments from the National Institute of Public Health of Kosovo's official Facebook page.
• Data collection spanned from March 12, 2020, to August 31, 2020.
• Comments were human-annotated in Albanian, a low-resource language.

Main Results:

• A comprehensive dataset of 10,132 Albanian COVID-19 comments was created.
• The dataset covers the initial phase and second wave of the pandemic in Kosovo.
• This resource addresses the scarcity of labeled data for Albanian social media.

Conclusions:

• The dataset provides a valuable resource for machine learning, information retrieval, and affective computing research.
• It enables deeper analysis of public opinion on health crises in low-resource language contexts.
• Facilitates evidence-based decision-making for public health agencies.