This project idea stems from our discussion we had with our Prof. Bing Liu about the difficulty people face on the web while zeroing on the right resource to understand a particular concept. Not able to find answer of questions through videos; which are easier to learn and understand. Many a times, the speech and images inside the videos go to waste and are never indexed by popular search engines. They are not included in search and hence user is not directed to them.

What it does

Our project crawls over the text in videos and generates keywords that we train it. We also crawl over slides discussed in videos and then we perform relevance search based on those keywords extracted.

How we built it

The whole system was built on top of Microsoft Azure. We used node.JS to code our script. We made use of Microsoft video indexer. Also we used Text Analytics API for video to make a judgement as to which video must be on top.

Challenges we ran into

The open source API did not give us a proper search experience. It took the order of words in exact form and also did not let us search through keywords, context, slides of video. So we developed a hack for that too. Hence, we were able to search efficiently.

Accomplishments that we're proud of

Video indexing of searches. The system putting in keywords by itself after being trained. Reading data from slides, whiteboards.

What I learned

I learnt Microsoft cognitive services, building npm packages, modifying APIs, deploying apps on Microsoft Azure

What's next for

Make a software driven video of our own which makes images from speech in continous form and it will form a video. Make a video rank algorithm for every subject matter. Then we will submit it for Microsoft Career Cup.

Where this project will be used in

This project will be useful in displaying targeted ads to online visitors of a particular website, which will show him/her the exact ads he/she might want to see based on their past activity

Built With

Share this project: