As AIMS Cameroon Coop Masters student, Sari’s project’s objective was to merge statistical models, such as ARIMA, which rely on linear techniques, with Deep Learning models that employ non-linear approaches. The aim was to develop Hybrid models that leverage the strengths of both linear and non-linear capabilities to extract valuable insights from time series data. It is worth noting that time series data can be dissected into both linear and non-linear components.
After graduating from AIMS Cameroon, Sari completed the AMMI program 2022/2023, his research project’s goal was to generate bitext data in the local language of Chad (Ngambay) and utilize this dataset to fine-tune pre-trained models. Due to the language being severely low-resource, a self-training approach was employed to generate synthetic data from monolingual data on the input source-side. The intention was to enhance the performance of the models by leveraging this technique.
He is currently doing a Summer Research Internship at EPFL. The project is about Modeling/Predicting Online Polarization using NLP techniques.