DostiNet

Inspiration

• Recent survey found that the majority of Americans have fewer friends than they did three decades ago.
• “Americans report having fewer close friendships than they once did, talking to their friends less often, and relying less on their friends for personal support.” by Survey Center on American Life.
• Per the shocking study, nearly half of all Americans — 49% — reported having fewer than three close friends. This marks a nearly twofold increase from 1990, when less than one-third (27%) of Americans had three or fewer close friends. - NYPost
• Naturally, the COVID-19 lockdown is partially to blame for today’s loneliness epidemic. According to the survey, “nearly half (47%) of Americans report having lost touch with at least a few friends over the past 12 months.”

What it does

DostiNet (Dosti दोस्ती means friendship in Hindi) helps you find friends based on your preferences and personality traits. These attributes are automatically assessed with Machine Learning and analyzed to find you friends with shared interest and background.

How we built it

• Database:
  • MongoDB
• Front End
  • HTML, CSS
  • Bootstrap
  • JavaScript
• Back End
  • Flask
  • Python
• Model Development
  • Python
  • spaCy for multi-dimensional word representations and semantic similarity

Challenges we ran into

• Defining semantic similarity in the context of our project
• Real-time update on each message

Accomplishments that we're proud of

• Deployed the backend on the GCP
• Achieved synchronicity between user messages
• Extracting keywords from chat history for users

What we learned

• JavaScript and DOM
• Natural Language Processing

What's next for DostiNet

• Include sentiment analysis for better friend-matching
• Ranking system of the user preference list
• Improve overall UI and user experience

Built With

• flask
• javascript
• mongodb
• python