Sentiment Analysis

Table of Content

• Overview
• Background and Motivation
• Goals
• Datasets
• Practical Applications
• Usage
• Milestones
• References

Overview

As Social Media start to influence the life of many children, teenagers, adults and now as we all know grandparents it can have a big impact on mental health. Those who have instagram, twitter, facebook, and tiktok hace been exposed to various posts that has affected us emotionally and psychologically. It can go from viewing a friends posts on travelling to different countries while your sitting at home doing work, if can be an invite that you never received, or any post that got you thinking "Their life is so much better than mine". Social media has also been a medium for spreading negativity, such as fake news, toxic comments, or even to ruin someones reputation by spreading rumours. However, it can't all be that bad? After all social media has allowed people to start their own business, create their own mood board, maybe even share their self-imporvement journey. What if social media can move in the direction of elevating one another?

You can access the web application through this link.

Background and Motivation

• Social media has influenced some people to think that their life is not interesting, they aren't as happy as they should be, they should be getting more "likes", have more followeres because if they don't then they are considered losers. All this develops anxiety and stress that can lead someone to become angry, depressed or feel extremly lonely.
• The rise on mental health for various social media users.

Goals

• Build a classifier to distinguish comments, posts, texts as negative, positive or neutral.
• Build a recommender system to feed into the individuals newsfeed positive posts such as images, inspirational quotes that elevates their mood.

Datasets

Prerequisite for downloading datasets from kaggle

• Have an account with Kaggle and create an API key (if you have not done this before) which will save as a kaggle.json file.

• Save your .json file under your home directory. If you are using Linux then it will go under ~/.kaggle/kaggle.json, if you're using Windos then follow the link here.

• Kaggle Sentiment140 Dataset - Good but not a lot of labels

• Kaggle Twitter Tweets Sentiment Dataset - Good but not a lot of labels

• OpenML Emotions--Sensor-Data-Set - This one is not bad

• Sentiment Analysis in Text - I like this one

• Emotions Dataset - I like this one

Practical Applications

• Social media apps tracking users in the background for mental health issues.
• Suggest accounts to follow that can feed the users newsfeed with positive ideas to prevent user from having negative thoughts that can cause them to doubt, or talk negatively about themselves.
• Integrated as a web extension to track users if they are experiencing any mental health issues.

Usage

• Clone Repo:

git clone https://github.com/kevinmfreire/sentiment-analysis.git

• Setup a virtual environment:

virtualenv .virtualenv/sentiment-analysis

• Activate virtual environment:

source .virtualenv/sentiment-analysis

• Install all requirements using pip:

pip install -r requirements.txt

• Download the dataset for this project:

kaggle datasets download -p ./data/raw/ --unzip yasserh/twitter-tweets-sentiment-dataset

• Setup the required NLTK packages

python nltk_setup.py

• To preprocess and train model you must be in the src/ directory and run the following:

python pre_process.py
python train.py

• To run web application stay in the main directory and run the command:

python app.py

Milestones

Phase 1:

Develope a webapp where the user inputs their sentence (email, tweet) we would return the label of that text (negative, positive, neutral sentiment)

Phase 2:

Talk to the Twitter API to gather the users latest tweets and analyze them (send them to the model) and return the users overall sentiment (mood).

Phase 3:

Include recommender system feeds the users to follow accounts that can imporve someones mental health.

References

• Text Classification Using Logistic Regression
• Text Classification using SVM and Naive Bayes
• Text Classification Using K-Nearest Neighbor