The problem Watch Wolf 🐺 solves

Track 4: Sentiment Analysis of social media posts

Note: Use

delta

and

cyLtEHTnwWw3WvphJVVT

as username and password for the demo.

The Problem

Social media has become a platform to spread information and opinions on a scale like never before. Sometimes, these opinions can take the form of bullying, infiltrate people's minds and manipulate them to execute counterproductive acts. However, drawing the line between the destructive and the harmless is difficult. And doing so manually on the scale that social media operates is impossible.

Note: Use

delta

and

cyLtEHTnwWw3WvphJVVT

as username and password for the demo.

Our Solution

Watch Wolf performs unopinionated multilingual sentiment analysis on text, images, social media posts and files. It provides an API that can scale up to a global level by leveraging cloud infrastructure. Watch wolf can identify and flag potentially harmful content for moderation.

Challenges we ran into

Note: Use

delta

and

cyLtEHTnwWw3WvphJVVT

as username and password for the demo.

Frontend

Since the API and Frontend development happened asynchronously, we faced issues while integrating with the API. We realised that using Axios over the fetch API provided a higher level of abstraction, reducing the integration time considerably.

Backend

We wanted a credential sharing system (for storing AWS and Twitter API keys), so deployments can happen in a platform-agnostic way. Instead of writing a custom config file reader which may have to handle POSIX and Windows file systems separately, we leveraged the existing AWS APIs. We resued the AWS profile file standard (traditionally used for storing AWS credentials) to keep our Twitter credentials, reducing the exposure to bugs that a self-written code might have.