Recently a video popped in my recommendation, this was the video - link. This video caught my attention and I watched it to the end. It got my mind thinking, I browsed the Internet and came to know that a machine called a Polygraph can actually detect truths and lies using body vitals. After further research, I found there are web tools that can spot lies using facial and audio recognition, Soon I thought we could make a tool that detects a lie by the text provided, it seemed to be a great idea. So we started this project

What it does

This is a text-based lie detector. Users can input text/statements and submit them. Our neural network will analyze the statement and classify it as truth or lie based on these categories

  1. No. of characters
  2. No. of words
  3. Avg. length of words
  4. Readability score
  5. Part of speech: Is this word an adjective, noun, something else After analyzing it properly it would tell you whether it is truth or lie and the truth/lie percentage as well. This tool can really help in spotting lies

How we built it

We used Flask and Python in the backend to create the web server. For detecting lies, we use a Neural network and NLP for our model. The model is trained on 200 rows of data by PolitiFacts, PolitiFacts is a website which provides facts and rates them as true or false. We have used CSS and vanilla JS for the frontend.

Challenges we ran into

We are both high schoolers and this project was a bit complex for us. Having limited experience in ML as well as neural networks we faced hurdles every minute. Using helps from open source platforms asking some friends and crash coursing things we made it through. But there were specific challenges that had us scared -

  1. Data Scraping
  2. Using NLP and Neural network (first timers)
  3. Our teammates left in between which only left us two
  4. Hosting it

Accomplishments that we're proud of

This hackathon was very competitive. Jostling through the competitive coding experience had us goosebumps. During these 24 hours, we achieved many things which made us proud and boosted our self-esteem. The accomplishments we are most proud of -:

  1. Completing this grand project in 24 hours
  2. Building the neural network
  3. Successfully data scraping over 800 rows (although we didn't use it at the end)
  4. Training the Neural network (was a too tedious task)

What we learned

We are both high schoolers and this project was a bit complex for us. Having limited experience in ML. We learned the basics of ML, expanded our knowledge of python, and learned about neural networks

What's next for Detectia

We will make it more accurate and provide reasons as well. We will also add OCR (image to text) and Speech to the text which will make it more accessible

Share this project: