Training Word2Vec model for emojis using twitter data.

We have 1 million sentences with emojis. This dataset is from twitter and you can download it from here.

1. Install requirements

   pip3 install -r requirements.txt

2. Training word2vec on our data

   python3 src.train_word2vec.py --data_path="Data/corpus.txt"

• This will train word2vec and save trained models as word2vec.bin

3. Create instance of prediction (In testing_emoji_model.ipynb)

   from src.prediction import Prediction
   predictor = Prediction(model_path="word2vec.bin")

   Once we have prediction instance we can use following methods.

   Methods in prediction

   1. getPrediction
   • Get only emojis for given text from top 200 most similar embeddings.

     predictor.getPrediction("cat",emoji_only=True, topn=200)

   2. get_similarity
   • How similar given two words are

     predictor.get_similarity(w1="😘", w2="😙")

   3. get_vector_embedding

      predictor.get_vector_embedding("🐰")

Pre trained model can be found here, Where woed2vec4 is final model.

1. Prediction

   

2. Similarity:-