This project explores developing a Word2Vec model for NLP using the Climate Fever dataset from the Hugging Face AI community. The Word2Vec model is trained to generate an embedding matrix, and its spatial relationships are investigated using cosine similarity. The quality of the embeddings are examined by exploring arithmetic relationships between similar and dissimilar words. The model is then compared with two pretrained models (1 Word2Vec using Google News 300 dataset, 1 GloVe trained using a combination of a Wikipedia dataset and the Gigaword dataset). All models are compared to evaluate the quality of embeddings within their respective matrices.
Nasa + Scientist = Laboratory
