faisalmehmood2007 / airline_sentiment_analysis

This repository performs sentiment analysis on tweets of an airline company

Airline_sentiment_analysis

This repository performs sentiment analysis on tweets of an airline company by using RNNs. The idea is to know how a user is feeling based on his tweet.

Requirements

• tensorfow
• numpy
• pandas
• scikit-learn

You can install all the requirements with the bellow command
pip install -r requirements.txt

Download pretrained word embedding

Before doing anything you should first download the pretrained word_embeding from GloVe by running the following commands

• clone the repository
• cd airline_sentiment_analysis
• run-- > python Download_word_embedding.py

Run inference

• clone the repository
• cd airline_sentiment_analysis 1- Run Through terminal
• open inference.py edit the text variable
• got to terminal and run --> python inference.py
• This will tell you the sentiment of the sentence 2- Run Through jupyter notebook
• open inference.ipynb
• edit text variable
• Run the cells to get the sentiment

Run training and evaluation

• clone the repository
• cd airline_sentiment_analysis
• open Train_and_evaluate.ipynb
• edit Dataset path and word embeding path if neccessary. (you won't have to edit anything if you kept folders as is)
• Run the cells

pipeline

1- explore the dataset:

• Envistage the dataset and knew that the data is skewed( more the 50% of the data is one negative and the rest is either nuetral or positive). The steps of the evistagation can be found in this notbook explore_dataset.ipynb

2- word2vec:

• Used a pretrained word embedding to converts words into vectors that represent each word

3- Model:

• Built a model consisting of three layers of GRUs followed by on dense layer

4- Training:

• Because the data skewed I implemented and used the Stratsified batch.

Accuracy, percision,recall , f1_score and a confusion matrix are all found in this notebook Train_and_evaluate.ipynb after training.