SentimentAnalysisTwitter
The assignment
This was an assignment in ML2 097200 course in Spring 2022 in The Technion – Israel Institute of Technology .
The field of detecting which emotion is represented in a text is developing and being studied due to its usefulness. For example, detecting if a review is positive or negative and more.
The data
The data is a csv file containing tweets and their labels according to the emotion – {happiness, sadness, neutral}. Every row in the file (except for the header) is an example.
Examples: (Notepad++ view)
happiness,Welcome @doeko ! Really glad to know you here. Your products rox man sadness,Disappointment really sucks! I'm getting used to it. neutral,I just want to Sleep.
The pdf report includes:
- Data loading and pre processing of the data.
- Architecture
- Loss function
- Optimization
- Regularization
- Hyper parameters tuning
- Train + Test loss graph
- Train + Test confusion matrix (a 3x3 matrix real x predicted)
- Train + Test accuracy
- Conclusions
LSTM Architecture
Glove embedding
Interesting results:
