YouTora (Unsupervised Baby)

• 

  searching videos by "literally"

• 

  searching videos by "impostor"

• 

  searching videos by"catastrophic"

Inspiration

We learn a language better when the messages(audio-visual contexts) are given, than when merely the meaning is given.

What it does

Given a word input, the website searches YouTube videos by captions (including auto-generated ones).

How I built it

Since YouTube api does not allow people to search videos by captions, I had to find a work-around. The way I did it is I downloaded captions of videos using API, and migrated them into a local database.

Challenges I ran into

• Django pagination, duplicates in database, search algorithm; tricky too build, and hard to debug. took most of my time during the hackathon.

Accomplishments that I'm proud of

• Finally I've made something useful!

What I learned

• I can widen my insight by asking questions to others; I asked questions about my project to the people around me during the hackathon, and they told me:
  • It can also be very helpful for visually impaired people
  • Could be great for collecting marketing data
  • Location-based statistics could give you some interesting insights
  • Try out other languages! learning Chinese and Korean is in high demand in the UK.

What's next for YouTora (Unsupervised Baby)

• Textual analysis: word cloud before and after the word, sentiment analysis
• deploying on the web with microsoft Azure
• web crawler which crawls around Youtube videos to collect video captions on the web.

Built With

• django
• python
• youtube