3 ms·
I run a similar service https://lettergram.net https://lettergram.net I like this layout - in terms of it being straight forward and easy to scroll. I am curi
by lettergram 8y ago
I run a similar service https://lettergram.net https://lettergram.net
I like this layout - in terms of it being straight forward and easy to scroll.
I am curious why some of the stories are trending (for instance (5 sources) at the bottom:
> 1,000 wild horses to be rounded up in Northern California (seattlepi.com)
It appears this is just following / aggregating standard news outlets (I'm guessing ~50 or so). Then clustering the stories based on some sort of fuzzy matching.
Pretty cool IMO, but care to elaborate on the machine learning portion?
- distilpost 8y agoIf you notice on the page, there are references to remarks made by actual people. The NER feature uses machine learning. Also the continuous clustering employs K-means and not just fuzzy matching as you stated. It is hard to imagine getting anywhere close to the level of accuracy with just fuzzy matching.