With the advent of social media news spreads at an unimaginable pace. Gone are the days when one had to wait for the morning newspaper to get the latest bulletins as news is now received at the instance it’s made. Moreover, individuals around the world have become reporters; sharing news of what’s happening in their communities at the moment it happens

However, with this power has come an unfortunate side effect. Purposely created misinformed reports also known as fake news are growing phenomenon. Sometimes this type of news is harmless and made for fun, other times it causes harm.

The Social Truth app plans on solving this.

What it does

The app allows users to fact check viral images they have received and verify the information they contain by simply long pressing their phones home button when viewing the image.The app works by scanning the user’s screen and searching for images which have been flagged as false. The app will show them a result if the image contains false information. The result will contain a link to the source and also a summary of the source explaining why the information in image is false. Users can then share the results with their friends to inform them of the false information. The app works in any app be it slack, discord, whatapp, twitter etc.

How False Images are added

The images to be flagged as false are added within the app by reliable people such as journalist. They simple upload the false image to flag it and a provide a link to a source explaining why the image contains false information. From the app will flag that as false [image]

Technical Explanation

The app uses the following services: AWS Lambda, AWS Rekognition, AWS Api Gateway and AWS DynamoDB. The model used is the DeepInsights Text Summarizer by Mphasis.

Technical Explanation of How Flagged Images Are Added

In order for an image to be flagged as false, a journalist has to upload the image to flag as false and as well as a url to the source. From there the app will extract the text content of the web page and send that to there server as a payload. The payload will also contain a image in base64 format

The payload is sent to an AWS Lambda function via AWS API Gateway. The Lambda function then calls AWS Rekogntion to extract the text from the image and it also calls the DeepInsights Text Summarizer by Mphasis model endpoint to summarise the text from the web page.

The result of the Rekognition OCR and the DeepInsights Text Summarizer by Mphasis are then saved to a AWS DynamoDB.

Technical Explanation of How Flagged Images Are Found

For an image to be flagged as false the following happens. The user activates the app by long pressing their home button or assistant button. From there the app will take a screenshoot of the current user's screen. The app then sends the image as a payload to an AWS Lambda function via AWS API Gateway. The Lambda function then calls AWS Rekognition to extract the text from the image. The text from the image is then searched against the existing flagged image texts on the AWS DynamoDB. If a match is found, the app will return the url source and the saved summary of the source.

What I learned

How the AWS Machine Learning market place makes life easy. With just a few clicks on has a full working model which they can use

What's next for Social Truth

We'll see after the Hackerthon

Share this project:

Updates