Tweets Clustering
Grouping tweets by content using K-means with Jaccard distance, an unsupervised take on topic discovery.
Overview
An implementation of K-means clustering over tweets using Jaccard distance between token sets, used to group short texts into topics without labels.
- Year
- 2022
- Area
- AI & Vision
- Role
- Machine learning
Stack
- Python
- K-means
- NLP
01
Custom distance
Jaccard distance over token sets, which suits short, noisy text.
02
K-means from scratch
Centroid selection, assignment and convergence implemented directly.