3 ms·
Combing 10k Hacker News Posts with Text Clustering
- mikewarot 4y agoClustering things conceptually near each other seems like an obvious thing to do. If my understanding of embedding is correct, it should be possible to take those embeddings at possibly pull things apart along other dimensions as you discover them. For example - the first chart "Top 10000 posts" seems (to me) to be oriented around these axis Programming/Technical Personal Distributed Deaths New Stuff Loss Future Finance / Stock Stuff Top 3000 Posts from Ask HN seems to be Self Improvement Personal Collective Coping with Reality Again, thanks for sharing with us a working set of tools to explore embeddings like this from a data source we're all familiar with.