Inspiration
Sentri was inspired by the looming concerns of the general public around social media. Over the years, social media discourse has evolved considerably. By using machine learning and data analytics techniques, we may be able to better understand the landscape of online discourse to better address societal needs.
What it does
Sentri searches public Twitter content for user-provided query terms. It processes the retrieved content and sends it to various Modzy APIs for processing. Sentiment, summary, and topic analyses are performed and results are indexed in the Sentri database. The indexed results are rendered for visualization for user interpretation.
How we built it
We used python, the flask web framework, sqlite, modzy, and tweepy.
Challenges we ran into
Building a fullstack app was challenging, coming from a data science and engineering background. There were also a number of challenges navigating the Modzy API, with occasional downtime and conservative rate limits. The latter was addressed by only sending 5 tweets at a time. However, we still have to wait for results to index, so in production there would need to be a redesign for more async behavior so users don't see blank visualizations if Sentri has not indexed particular content quite yet.
Accomplishments that we're proud of
It was an accomplishment to be able to use novel tooling in order to create a proof of concept of something relevant to current challenges.
What we learned
We learned that NLP is REALLY HARD for social media content!
What's next for Sentri
Refine and scale!
Log in or sign up for Devpost to join the conversation.