Learning Management Systems (LSM) have been adopted to provide interaction between instructors and students in online learning courses. In these environments, large volumes of messages are exchanged, making it difficult for the participants to follow all the discussions. When messages are ignored, students' opportunities to effectively construct knowledge through online discussions are reduced. In this context, this paper proposes a method for mining the content of messages written in Brazilian Portuguese to enhance students and instructors’ interactions in online discussion forums. Four different automatic methods were developed, one for each of the following four dimensions of the originality and collaboration processes: (i) Semantic textual similarity to measure originality; (ii) Expressing appreciation towards other participants; (iii) Recognizing group presence; (iv) Sharing information and resources; (v) Soliciting feedback/Answering questions. The use of the methods for analyzing online discussions forums was validated in a case study conducted in a real course. A discussion forum tool was developed to provide feedback for students and instructors about the originality and collaboration level during online discussions. Results showed the potential of the proposal in promoting students' collaboration, enhancing interaction and reducing plagiarism in discussion forums. André, M., Pinheiro, A., Rolim, V., Lins, R.D., Falcão, T.P., Mello, R.F. (2022) Adopting Learning Analytics to Promote Collaboration in Online Discussions Written in Portuguese. Computer-Based Learning in Context, 4 (1), 1-15.