DupeDetector

Inspiration

The popular documentary, The Social Dilemma, was a major inspiration for this project. With the rising use of The Internet, misinformation is extremely easy to spread. The documentary outlined the mental health impacts of social media and its rise in fake news. According to a MIT study, fake news spread 6x faster than real news on Twitter. Additionally, according to the CIGI, 86% of the world’s population has fallen victim to fake news. With an abundance of social media posts intending to ‘spread awareness’, we must fact-check it. That is where DupeDetector comes along. We created a website and connected our ML model. We used fake news datasets from kaggle to create our model.

What it does

DupeDetector is a website that detects if the user is consuming fake news or not. It allows the user to input text from the article they are reading, or an image of an article, tweet, or instagram post. The website will then extract the text from the image, or take the text and determine its reliability. There is also an option to view your input to see if our code has extracted your inputted text correctly.

How we built it

We built the ML model using Python, our frontend using HTML, JS and CSS, and our backend with Python and Flask.

Challenges we ran into

We wanted to try something we haven't done before, and decided to create a project that utilized machine learning. We spent the first day just generally watching introduction to machine learning videos and the next day discussed what kind of project we could create.

We also struggled with connecting our ML model to our flask and creating our ML model because it was our first time working with machine learning and Jupyter Notebook. We jumped a lot of hurdles when trying to debug our code because of the new functions we learned in our ML model.

Accomplishments that we're proud of

We are proud of creating a Jupyter Notebook Machine Learning project (with a 95% accuracy score!) for the first time and being able to connect it to a web app.

What we learned

Since it was our first time, we learned a lot about machine learning and connecting machine learning models to the backend of a website. We learned a plethora of new algorithms used in ML, such as a decision making tree and linear regression.

What's next for DupeDetector

Next, we hope to add more data in our model to improve its accuracy. The Kaggle Dataset was mainly American news, therefore we hope to add more data from other countries. Another good addition would be to determine the accuracy of news in languages other than English.

Built With

• css3
• flask
• html
• javascript
• jupyternotebook
• python