In today's Big Data era, relational databases and NoSQL databases coexist. The relational database is known for its limitation in terms of scalability. On the other hand, sentiment analysis is a Big Data product that uses unstructured documents such as review results, social media content, survey responses, etc. For this reason, this study will focus on
how to design and build a database that can handle thousands of documents containing opinions on social media. The database model built in this study focused on its conceptual design based on previous research frameworks. The proposed model will be implemented using one of the infamous document-oriented NoSQL database MongoDB then tested on a sentiment analysis application built under ShinyR framework. The results showed that when comparing the
sentiment analysis process with a lesser amount of data, it only affects two operation speeds in the database: insert and get more operations. The speed for insert operations in the analysis process using 18,900 documents is at 16,800 ms and the speed for get more operations is at 3,000 ms.