Please read me!!!

A project for Language Processing 2(Final Project) - JianWu(xcb479)

Note

ipython notebook demo_BOW.ipynb

or

ipython notebook demo_word2vec.ipynb

The quickly link for Result

• Final Result for BOW.
• Final Result for word2vec

About ./sentiment

I write all available funtion in ./sentiment

Credits

• fastText by facebook
• the keras tutorial on the subject
• word2vec

Example output of current model

A cinematic achievement of amazing depth. Marvellous acting, captivating score, a journey in a time when directors produced masterpieces. one of a kind, my favorite film, I LOVE IT!

gives 0.75

I went with my girlfriend to the cinema. She's into american comedies and I really like her, so I just saw the worst most terrible movie in my life. boring, untalented acting, dull and uninspired cast, absolutely terrible score. the worst movie ever.

gives 0.14

with Robert De Niro and Al Pacino

gives 0.45

with Adam Sandler and Jennifer Aniston

gives 0.38

Training

The data set background for training and testing.

Also

You can type

from sentiment.utilities import download_data

download_data()

to download dataset

The model for training:

• SVM
• Random Forest

Final

I gave up the LSTM since the absolute error was very big As the comments mentioned.

• Final Result for BOW.
• Final Result for word2vec