Jupyter notebook contains solutions of the following below task:
Given Reviews find positive or negative
- Developed a rule-based classifier with logistic regression, achieving 70.17% accuracy, and implemented a Bag-of-Words model that reached 85.02% accuracy, identifying the top five most positive and negative words.
Learn Dense word embedding and test its semantic and analogy.
- Trained a SkipGram model from scratch to learn word embeddings using the WikiText2 dataset. Evaluated the quality of the embeddings using an analogies dataset, achieving a precision-at-5 score of 40%