In the digital age, sentiment analysis plays a crucial role in processing large volumes of data for timely decision-making. This research introduces a novel integration of Robotic Process Automation (RPA) and Knowledge-based Systems (KBS) for sentiment analysis, specifically designed to address limitations in existing systems. Unlike traditional methods, which rely heavily on predefined rules or static keyword sets, this approach uses the PhoBERT model for advanced Vietnamese text processing and uses cosine similarity to classify data more accurately. By automating the collection and segmentation of learner feedback on social networks, the system achieves approximately 80% accuracy in classifying sentiment when deployed at several universities in Vietnam. The integration of RPA enhances the efficiency of the system by reducing manual intervention, while KBS ensures real-time updates to the knowledge base, overcoming the dependency on external databases. This new integration provides a scalable solution, improves data processing speed, and increases adaptability compared to traditional sentiment analysis models, helping higher educational institutions understand and engage with learners. These findings can be applied and further extended to a wide range of practical domains.