Emekadavid/South-African-Language-Identification

An ML model that predicts South African languages based on given text # South-African-Language-Identification An ML model that predicts South African languages based on given text # Project description This project is a hackathon organized by the Explore Datascience Academy for students. In this project, students are expected to build a model that would detect South African Languages. Eleven languages from the data set curated by the South African Ministry of Arts and Culture were given. The predictions made by the model were submitted on the Kaggle competition page. # Classifiers Used During The Project For the project, several classifiers were analysed. Some of them were MultinomialNB classifier, Logistic Regression, Ada Boost, Random Forest Classifier, Linear SVC and Bagging Classifier. The libraries for these classifiers were taken from scikit-learn. Scikit-learn is a popular machine learning library. Among all the classifiers, MultinomialNB was chosen as the classifier that gave the best f1-score. An f1-score calculated the average of the precision and recall of a model. It is a metric that is used to evaluate how well a model can predict an unseen data after training on a given set of data. # Challenges of the Project One of the challenges of the project is the time constraint. Students were given only 48 hours or less to come up with a good model. Also, I kept getting RAM errors on the Google Colab page I was using to tune the hyperparameters of some of the models. If not for these two challenges, I would have found a model that would give close to a 100 percent f1-score on Kaggle. Also, I believe that given the time constraint, I would have been able to try out a neural network classifier. After the competition, I would do that and improve on the model. # Conclusion Classification and detection of languages is an emerging field of study in natural language processing (NLP). I believe that the experience I received from this project will help me when I am working with an employer. I would like to use this opportuni …