TweetSOS

Inspiration

The inspiration behind the TweetSOS disaster response app was to leverage the power of social media and machine learning to assist in disaster response efforts. With the increasing use of social media platforms, valuable information related to disasters is often shared through tweets. We wanted to create a tool that could accurately analyze tweets in real-time and predict their relevance to disaster events. By doing so, we aimed to provide prompt support to emergency responders and aid organizations, contributing to sustainable development efforts and ensuring the well-being of affected communities.

What it does

TweetSOS is a disaster tweet predictor web application. It takes user-input tweets and uses a machine-learning model to determine whether the tweet is relevant to a disaster event or not. If the tweet is relevant to a disaster, the app provides information on how to contact the relevant authorities or organizations for disaster response. On the other hand, if the tweet is not related to a disaster, it indicates that there are no disasters mentioned in the tweet. The app also features an introduction to its purpose, key features, and a call-to-action encouraging users to actively contribute to disaster response efforts.

How we built it

We built TweetSOS using Python as the primary programming language. Machine learning models and libraries such as Pandas, NumPy, and Scikit-learn power the app's core functionality. The data processing and text cleaning was done using regular expressions and the NLTK library. We utilized the TfidfVectorizer to convert text data into a numerical format for the machine learning model. For the prediction, we used a Logistic Regression model, optimized using GridSearchCV for hyperparameter tuning. The app's user interface was developed using Streamlit. Streamlit allowed us to showcase the disaster tweet predictor and provide a user-friendly interface for entering tweets and viewing the results.

Challenges we ran into

Data Preprocessing: Cleaning and tokenizing tweets while preserving meaningful information was crucial for accurate predictions. Handling special characters and removing stop words posed some difficulties.

Accomplishments that we're proud of

Despite the challenges, we are proud to have developed a fully functional and user-friendly disaster tweet predictor app. The successful integration of machine learning algorithms and creating an interactive interface through Streamlit allowed us to achieve our goals of empowering disaster response efforts and contributing to sustainable development.

What we learned

Building TweetSOS taught us valuable lessons in various areas, including: Natural Language Processing (NLP): We gained experience in text processing and tokenization techniques for NLP tasks. Machine Learning: Selecting and fine-tuning machine learning models for specific tasks and understanding their limitations. Streamlit: Creating interactive web applications using Streamlit and customizing the user interface. Data Science Workflow: The end-to-end process of data collection, preprocessing, modeling, and deployment in a real-world application.

What's next for TweetSOS

Moving forward, we have several exciting plans for TweetSOS: Real-time Data Streaming: Integrate real-time data streaming to monitor and analyze tweets for immediate disaster response continuously. Multilingual Support: Enhance the app to handle tweets in multiple languages to broaden its usability. Geolocation Integration: Implement geolocation analysis to pinpoint the location of disaster-related tweets and assist in targeted disaster response. User Community and Feedback: Engage users and disaster response communities to gather feedback and improve the app's accuracy and functionality. Expand to Other Social Media Platforms: Extend the app's capabilities to analyze and predict relevant posts from other social media platforms, such as Facebook and Instagram. We believe that these enhancements will further strengthen TweetSOS's impact on disaster response efforts and contribute to the broader goals of sustainable development and well-being for all.

Built With

• python
• streamlit