Emotable: A Sentiment Analysis Model
- 1 minEmotable is a sentiment analysis model trained on tweets from the twitter API. It classifies tweets as positive or negative based on the tweet content.
I created the training and validation datasets by retrieving tweets from the Twitter API that had certain “positive” or “negative” emoji using a TweetCollector class that I built. I got the list of positive and negative emoji from an existing dataset . I made the assumption that a positive tweet is one that contains a positive emoji and no negative emoji and vice-versa for negative tweets.
To build the model I trained a Long Short Term Memory network on 80% of the data I collected and validated it on the remaining 20%. This was my first machine learning project so I did this by adapting a tutorial for creating a sentiment analysis model for movie reviews from IMDB data. The model achieved 85-90% accuracy on classifying tweets as positive or negative.
Languages/Technologies
- Numpy, Pandas: Python libraries for data analysis/manipulation
- TensorFlow, Keras: Python libraries for training/building ML models
- Twitter API: datasource
Quai Wentt
learning-obsessed software engineer