joyce-lin / project_twitter_nlp Goto Github PK

This is my capstone project for the Data Science Immersive at General Assembly.

In this project, my goals are:

1. Set up real-time Data Collection process and Data Infrastructure
2. Examine different Natural Language Processing tools on collected tweets
3. Create A|B Testing Model on similarity comparison
4. Use Time Series Modeling to catch the trends
5. Tuning hyperparameters for model improvement

To test my framework:

1. I collected and cleaned over 1.5 million tweets from uing TwitterStream API

      /lib/get_tweets.py
   

2. Create scheduled and on demand LSA processing for text vectrozation

      /ipynb/01_Fit_pipeline_TfiDf_SVD.ipynb
   

3. Event and Trend Detection using Cosine Similarity and ARIMA modeling

      Event extracting using TFIDF and SVD
        /ipynb/03_Tweets_Modeling_CosineSim_AB_Test_SVD.ipynb
      
      Hashtag Time Series Modeling 
        /ipynb/05_Hashtags_Modeling_WhatsTrending.ipynb
   

1. Python
2. TwitterStream API

1. postgres
2. redis

1. NLP (LSA): SAPCY | TFIDF |SVD | Count Vectorizer
2. Cosine Similarity
3. ARIMA Modeling