Tweetailyze

The front-end for Tweetailyze, a webapp that summarizes a Twitter account's tweets through embedding/clustering of tweets, w/ sentiment analysis and topic modeling.

How it works

1. POSTs a Twitter account's username to the backend.
2. Gets the most recent X tweets using the tweepy package.
3. The text of each tweet is preprocessed using the nltk package.
4. The preprocessed text is transformed into embeddings using OpenAI model 'text-embedding-ada-002'
5. The embeddings are clustered using agglomerative clustering with cosine metric and average linkage.
6. The tweets are separated into clusters based on their cluster labels.
7. Sentiment analysis is performed using the VADER sentiment analyzer.
8. Topic modeling is performed using non-negative matrix factorization (NMF) from the scikit-learn package.
9. The top words for each topic are classified using the OpenAI model 'text-curie-001'.
10. Results are stored to Postgres and returned to the frontend.