Recommender System

In this project, I wrote an app that takes topic and sub-topic and time T as input data and returns the most relevant streaming tweets to the topic and sub topic in the last T seconds. The streaming tweets are collected, cleaned and vectorized to get the result.

Here are a list of descriptions for each folder in this reposition.

• jupyter notebooks/twitter_proj.ipynb

  A notebook that documents the steps

  • save streaming related tweets to the topic in the given time frame
  • clean and pre-process the saved tweets
  • vectorization using the pretrained model GoogleNews-vectors-negative300.bin
  • return the most similar tweets related to the given topic and sub-topic by means of the similarity score

• Twitter/scripts/lib

  contains

  • data_collection.py

    gets the latest tweets for the topic in the time frame

  • data_preprocessing.py

    cleans and vectorizes given tweets

  • generate_recomendations.py

    generates the most related tweets to the user input topic and sub-topic

• app.py

  a Flask app that user submits Sub-Topic, Topic and Timeframe at /user_input.html and it returns the most relevant tweets at /results.html

• Twitter/scripts/front-end

  • /user_input.html

    a table for the user to submit Sub-Topic, Topic and Timeframe

  • /results.html

    a table of results of the related tweets is saved here.