Hashtag Recommender October 2013

Project Description

The project users Twitter API to suggest mentions and hashtags based on multiple criteria. Demo Link. Users can find relevant twitter hashtags and mentions based on the following:

  • Friends: Recently used hashtags and mentions from your home line (people you follow)
  • Nearby: Recently used hashtags and mentions within 3 miles of your current location
  • Specific Users: Recently used hashtags and mentions for a specific user. TO use, type a username in the left username box anc click "Grab'em".
  • Suggestions: Recently used hashtags based on what you type in your tweet. To use, type your tweet and click "Suggest Tags". An algorithm will analyze the text and suggest mentions and hashtags. Note: Multiword matches are ranked higher than single word matches.

Team Members

  • Hassan Jannah: UI and Backend object classes
  • Ramit Malhotra: Text analysis & backend
  • Christopher Fan: API calls, PHP, & ranking function

Technical Description

The application uses ability of Twitter to generate a list of suggested hashtags based on the text body of a Tweet. Here is an example from 12/26/2013. User enters text for a tweet: “The new scorcese movie is great”. The application filters out commonly occurring words. Using the Twitter API, searches are conducted for the remained keywords. Generate a histogram of popular hashtags based on the frequencies of hashtags returned from the search queries. The most popular hashtags are the most relevant.

The hashtag recommendation is able to recognize topical keywords like proper nouns very well. Also, for recently occurring events, the recommendation system is able to make inferences. In this example on 12/26/2013, the tweet body is Based on that date, the recommendation system recognizes the keyword “scorcese” as being highly relevant and also correlates to the movie “#wolfonwallsteeet” because on that date, the movie is related to the keyword “scorcese”.

Github link: https://github.com/christopherfan/metacrap