<![CDATA[The current development of information technology has increased the dynamics of public opinion regarding the direction of education policy in Indonesia, which is spread across various digital platforms such as social media and news portals. The linguistic diversity used by the public, including Indonesian, regional languages, and mixed foreign languages, poses a challenge for conventional sentiment analysis, which is generally monolingual. This study aims to implement the Multilingual Sentiment Method to analyze and map public sentiment perspectives toward education policies in Indonesia more comprehensively and accurately. The research methodology encompasses data collection utilizing web scraping and APIs, data preprocessing (cleaning, case folding, tokenization, stopword removal, and normalization), automatic labeling, and the application of Natural Language Processing (NLP) using multilingual pre-trained models. Model performance evaluation is conducted using accuracy, distribution, precision, and prediction metrics. Data visualization analysis is also performed to illustrate sentiment distribution and emerging issue trends. The results indicate that the multilingual approach successfully improves classification performance compared to monolingual methods, particularly in detecting sentiments in mixed-language texts. The distribution of public sentiment is dominated by neutral and positive categories, whereas negative sentiments emerge regarding issues of technical implementation and equitable access to education. In conclusion, the implementation of the Multilingual Sentiment Method is effective in providing real-time mapping of public opinion, thereby supporting the formulation of education policies that are more responsive and data-driven.]]>
Copyrights © 2026
