The video demo shows what needs to be done. The steps that were used to make the demo are:
Plug the phone that shows the question live into my MacBook and use QuickTime to have the screen appear live on the computer.
Use Automator to run a script after taking a screenshot of the question and the 3 answers.
Upload the screenshot to Google’s Vision API and use the Text Detection to read the question and the 3 answers.
Use a Google Custom Search Engine API and enter in the question. You can also use Boolean Operators or quotes surrounding the answers at the end of the question.
Scan the 9 results this API returns and give each answer a score based on how many times it occurs in the snippets.
If the answer cannot be found in the snippets; load the websites one by one, search all of the text and give each answer an occurrence score.
You can choose to do it this way or any way as long as it achieves the desired result. It is crucial that it is all automatically done after the user has taken the screenshot and done in around 7 seconds. Let me know if you can do this.