sentiment-analysis
this is the repository for my graduation project: sentiment analysis on twitter for brands and corporations. twitter mentions for airline companies are used as the training data.
the model uses word embeddings and a bidirectional LSTM neural network to predict the sentiment (positive, negative, neutral) of a tweet/customer feedback.
further improvements will include using part-of-speech tags and dependency information of words in addition to word embeddings.
current test accuracy score: 81%
dataset sources:
You need to download the dataset before running the code.
How to run:
- create a new conda environment and activate:
conda create -n python=3.6
conda activate
- install requirements
python3 -m pip install -r requirements.txt
- create word embeddings
python3 word_embeddings.py
- create and evaluate model
python3 sentiment.py