The problem Stealth AI solves

##Inspiration
Everyday people watch one billion hours of videos on YouTube and generate billions of views (YouTube, 2019). Another interesting fact is that searches related to the term “how to” are growing 70 percent year over year, according to searchengineland.com. This means that when users are searching for ways to learn something, they rely on video content. In the limits of COVID-19, the whole world realizes the power of online education and the impact it can have on millions of people.

##Problems

Many times we find ourselves skipping through a YouTube timeline trying to find a punchline or a specific part of a tutorial. It can be time-consuming and mind-numbing to try and find a specific moment in a YouTube video. (And this doesn’t end after one or two vids)​

Although plenty of video solutions available online and usage is extravagant, often these lectures can be too long and monotonous. People zone out frequently and lose track of what is going on and end up rewinding the video and watching again and again. ​ ​Also, in this process there is a lack of feedback and there is no one to test you if you have really understood what you are listening to.

The transition to online classes has, contrary to expectations, increased the average time spent for classes. To improve studying efficiency, we decided to tackle a major part of our long study times and comprehension ability: Zoom lectures. We thought it’d be nice if we could somehow condense lecture material without losing comprehension and possibly even increasing our understanding.

Challenges we ran into

##How we built it

Youtube Video Analyser: For the youtube analyzer, we first generate the summary of the available youtube transcripts/closed captions using nltk summarizer, which is done on the azure function which we have developed. This azure function returns the summary and transcript of the youtube video to the web app. With the help of this summary of the video, we generate the keynotes from the entities API, highlight key main sentences in the summary, showing the main topic using the relevant API. We also do the sentimental analysis of the youtube latest comments using the sentiment API. For the quiz generation, we use the disambiguation API which gives the sentences with the tokens which further uses the azure function from where we get similar words for the options of the question. API Used: /relevents /sentiments /entities /disambiguation

Zoom Online Lectures Analyser: Here we used the zoom video transcripts and led by the text cleaning and generating summary with the previous summarizer azure function. After generating the summary we used the expert.ai APIs for generating the key elements, phrase searching, voice searching, highlighting the main sentences and generating the smart quiz as mentioned above. API Used: /relevents /sentiments /entities /disambiguation

Legal Document Analyser: For a legal document analyser we read any privacy policy from the text file and generate the summary of it and use the relevant API for the main sentences and main lemmas for highlighting the important information of any legal document. API Used: /relevents
Youtube Video Analyser: We find ourselves skipping through a YouTube timeline trying to find a punchline or a specific part of a tutorial and end up spending a lot of time. To remove this hassle we have created a youtube video analyser, where we can get the key moments, summary, searching through phrases, voice searching by just putting the youtube video link.