As students from diverse backgrounds, especially underrepresented minorities, sometimes we need someone to look up to. We want to relate to the conversations happening among industry professionals on Twitter. Moreover, we want to enhance the diversity of perspectives shared in a conversation thread on the platform. We believe there are many voices that are unheard and unexpressed simply because they feel like they do not belong. This features aims to show everyone that we all belong.
What it does
After using NLP to analyze the most relevant and usual topics of conversation from a user timeline on the Twitter platform, it provides a search based on that topic filters the tweets by the most uncommon ethnicity group engaged in the conversation (based on users name and bio ideally). Then gets that tweet id and pushes a notification to the user who belong to the most underrepresented ethnicity and is mostly tweeting about that subject. This encourages users to engage in conversations otherwise the would not be aware of, diversifying the conversations
How I built it
- Figma for design
- React for front-end
- Flask for back-end
- Tweepy, a Twitter API wrapper to obtain tweets that were then stored in Firebase via Google Cloud Platform
- The Google Cloud Natural Language API was used to classify the tweets by topic using text analysis
- And we used another NLP API that classified users by ethnicity from their name
- Identifying threads --- Classify by topic --- Determine distribution of backgrounds of users participating on thread - if skewed: this thread will be shared with diverse professionals
- Identifying diverse professionals --- Determine background or expertise by running NLP on recent 50 tweets --- Infer ethnicity from name
- Match threads to diverse professionals
Challenges I ran into
Connecting React with Flask. Running the Firebase classifier. Running the classifier in many tweets without breaking the algorithm or running into errors. Teamwork on different time zones.
Accomplishments that I'm proud of
What I learned
Using Twitter API, Firebase, Team Work. NLP API. React and connecting it to Flask
What's next for Diverse Voices
Adapt to Android and iOS. Better NLP models to classify tweets and users.