Question1_SVD : Implement a Word embedding model and train your own word vectors using Co-Occurrence Matrix by applying Singular Value Decomposition (SVD).
- filename : 2020201092_NLP_Q1
- embeddings : weight_matrix_q1.pkl- https://drive.google.com/file/d/1QZWKZrN2r9Ey78pSm4LPhQSOgAPBKyI1/view?usp=sharing
to get the embeddings for my model we need to get it from the link mentioned above under the name embeddings_q1.pkl and to load it into a variable
Instead of using co-occurence matrix, I have used DOK matrix..it is an implementation of co-occurence matrix but with faster computations and less memory.
run the command ->
import pickle
data = open ('path_to_folder/weight_matrix_q1.pkl', "rb")
embeddings = pickle.load(data)
and then we can use the embeddings to further work
Top 10 words and closest words to camera and the tsne and scatter plot and comparison with the embeddings of gensim are in both the ipynb file for both svd model.
http://snap.stanford.edu/data/amazon/productGraph/categoryFiles/reviews_Electronics_5.json.gz