Inspiration
After seeing the power of IBM Watson sentiment analysis, we saw that it could be as a signal for recommendations engines, and since we already had YouTube in mind, we thought of applying the Sentiment Analysis service to youtube recommendations.
What it does
It monitors user’s twitter account. Whenever the user posts a tweet, the system gives the tweet a sentiment score. From that we could determine the emotion state of the user and suggest appropriate youtube content to the user. For example, when we know the user is depressed, we may filter out the videos that may have negative impact to the sentiment of the user.
How I built it
We first get the user’s sentiment from monitor what kind of tweet the user posts. We then get a list of videos from the user’s youtube homepage (Which has recommendation videos from youtube). We then get the captions of each video on the page, run the sentiment analysis to determine whether this video should appear on user’s screen depending on the user’s emotional state. The category (or genre) of the video would have impact on the decision making process of the system as well. If the genre is comedy, then it will have a higher chance to classify it as a positive video.
Challenges I ran into
We have faced a lot of challenges. The biggest problem is to come up with a project idea. We also had trouble hooking up with different kinds of APIs mostly Youtube, Bluemix. And Javascript even-driven paradigm was new challenge to us.
Accomplishments that I'm proud of
That we pushed through the whole night to finish the project, without giving up.
What I learned
Getting more familiar with APIs from Watson and Youtube. A perfect time to review our javascript skills.
What's next for Morphy
Can also use Facebook post to provide a higher accuracy on analyzing user’s sentiment. We can show quotes on the screen depending on the user’s sentiment. Or offer some consolation and advice when the user is feeling down.
Log in or sign up for Devpost to join the conversation.