Indobert Sentimen Chromebook

IndoBERT Sentimen Chromebook is a fine-tuned Indonesian language sentiment analysis model based on IndoBERT, designed to analyze public discourse surrounding the procurement of Chromebooks. The model employs a multi-task classification architecture with three distinct heads, each addressing a separate classification task. The first head detects spam, distinguishing between spam and non-spam comments, as well as identifying comments related to gambling or promotions. The second head classifies the stance of comments, assigning labels such as pro-Nadiem, kontra-Nadiem, kritik_peradilan, kritik_pemerintah, netral_informasional, and tak_jelas to indicate the position of a comment toward specific actors or issues. The third head identifies the emotional tone of comments, with possible categories including marah (anger), duka (grief), sinis (cynicism), harapan (hope), and netral (neutral). Training data for the model consists of 6,233 Instagram comments related to the Chromebook procurement case, spanning 11 posts and covering a period from September 2025 to July 2026. Of these, 321 comments have gold labels and 5,912 have silver labels. The model's performance is reported with a macro-F1 score of 0.507 on the stance classification task using the gold validation set. IndoBERT Sentimen Chromebook is implemented in PyTorch and is available under the MIT license, making it accessible for use and modification. It is particularly suited for analyzing Indonesian-language social media commentary in the context of public sector procurement discussions. The model is hosted on the Hugging Face platform, where it can be accessed for text classification tasks involving sentiment, stance, and spam detection in Indonesian.