We, Amer Elsheikh and Abdallah Abdelaziz, designed and implemented a fully working ML model to predict tweets' sentiment with a 79% accuracy as a group project in Fall 2021. We first made a literature review and chose sentiment140 (1.6M tweets) dataset to work on, then we preprocessed the data and generated useful featurs. Then, we experimented different Sklearn models on the data, and the best was a neural network model associated with a glove-twitter-200 word embedding based on context, which gave 76% accuracy. Thus, we decided to implement a neural network from scratch but with our own word embedding as an embedding layer part of the network (based on sentiment) which gave 79% accuracy. Lastly, we made a simple Django website for users to interactevly use the model. It also supports online learning, where we ask the users to provide feedback for the prediction and retrain the model using new data every
-
Notifications
You must be signed in to change notification settings - Fork 0
QEDady/Twitter-Sentiment-Analysis
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
A Neural Network with an Embedding Layer made in python from scratch , no ML libraries used, to predict tweets' sentiment with 79% accuracy.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published