balksi

7 ms·

1. 1.▲
   by balksi 10mo ago

   For similarity it is important to consider the dimensionality of your embeddings. The larger the text you wish to compare the bigger each embedding should be (to my limited understanding). So a paragraph might be good as a 384-dim vector bu
2. 2.▲
   by balksi 10mo ago

   I am working on yet another news aggregator: newsmuncher.com So far i've got the scraping and embeddings / similarity clustering down (to build timelines of news stories), lots of data cleaning and UI refinement required. I find i
3. 3.▲
   by balksi 1y ago

   +1 calendly solved these issues for me