Readme v češtině na tomto odkazu
A project where artificial intelligence is used to distinguish between fake and real information. 📰
- Clone repo to your local machine with
git cloneand link from github - Open directory and start
install.batfile. This will install all necessary libraries and also will start a program And it is done. Your web server and python api is running.
In this simple diagram you can see how the whole program works.
First, we receive URL on web server. We take this url and pass it to scrapePage.cjs. Function of this program is described deeper in this readme.
Then, we send post request to python api containing scraped text as body. Python API gets this text and calls Ai model to decide truthfulness of text. When decision is made, we send it back to web server and it shows on frontend.
There is a python AI model, trained on a dataset from kaggle. It uses Sklearn and Pandas 🐼 library. We use function classify_text, which we then export to file server.py. In server we use flask library for own API. API takes text as body and returns true/false based on prediction about the news. Dataset used is from some Kaggle project. Accuracy of this Ai model is 0.92230, meaning it is trained and 92,23% same as data that we have provided.
We used a javascript library called Pupeteer to fetch an h1 element from a given website, after doing so, the program returns the result and sends it over to our Python API that decides whether the information is according to reality or is but a fabrication.
We created a simple, yet stylish frontend for our application, adding a monochrome design to it.
| Color | Hex |
|---|---|
| Starry night | |
| Midnight | |
| Snowy white | |
| Elephant tusk |
| Font | Link |
|---|---|
| Consolas | https://en.wikipedia.org/wiki/Consolas |
We added a simple solution, to check, whether the input is a valid url.
