Skip to content

A machine learning model which takes a sentence and finds the most appropriate emoji to be used with this sentence to make the text messages more expressive

License

Notifications You must be signed in to change notification settings

parasgulati8/Emojifier

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Emojifier

Thisproject aims at developing a model which inputs a sentence and finds the most appropriate emoji to be used with this sentence to make the text messages more expressive.

There are two versions of the Emojifier. First version uses the averaged values of all word vectors in a sentence and then using softmax function for classification.

Second method uses the LSTM layers to learn the long range dependencies in the text and thus performs better than the first model.

Word Vector Representations

We use pretrained 50-dimensional GloVe embeddings to represent words in vector form in a 50 dimensional space. The word embeddings were taken from Stanford open source project : https://nlp.stanford.edu/projects/glove/.

Conclusion

  • Emojifier is an application that inputs a sentence and (such as "Let's go see the baseball game tonight!") and finds the most appropriate emoji to be used with this sentence (⚾️).
  • It uses GloVe representation to turn words into 50-dimensional vectors which are then input to the model for training and prediction.
  • Pretrained GloVe embeddings are used and the weights were not changed while training the LSTM model.
  • The LSTM model achieves an accuracy of 82% on test set

References

The project is a part of an online course on Sequence Models that I completed on Coursera

https://www.coursera.org/learn/nlp-sequence-models

About

A machine learning model which takes a sentence and finds the most appropriate emoji to be used with this sentence to make the text messages more expressive

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published