Kaggle Silver Medal - Natural Language Processing - Curriculum Recommendations
- Achieved a Kaggle Silver medal: 41/1057 (top 4%)- Natural Language Processing task on providing high-quality course recommendations for a learning platform The Learning Agency Lab- Created special tokens and used them to construct the model's input- Used a multilingual Sentence-Transformer model with language abilities, such as AIDA-UPM/mstsb-paraphrase-multilingual-mpnet-base-v2, for Stage 1 training- Used the kNN method of the RAPIDS library to retrieve the most similar 100 content embeddings for each topic, based on Stage 1 embeddings- Trained the Stage 2 Cross-Encoder model using the Stage 1 model as a base, and adjusted the classification threshold value based on the F2-score