Re:Show HN: Toxicity Analysis of Twitter Handles
by dekervin on Sunday March 07, 2021 @11:00AM
(#26376885)
Good job! although it would be really useful if you reduce the computational load and then simply analyse the last10 tweets. Itwould be more useful.
[ Parent
| on HN ]
Re:Show HN: Toxicity Analysis of Twitter Handles
by 9notorp on Sunday March 07, 2021 @11:38PM
(#26382325)
Thank you! Not sure if I can do that unless browsers go back to the old designs and make ipc possible among tabs/processes again, the compute happens on the browser so it entirely depends on the end user hardware config. And while making this I gave that approach a try as well but decided against it, since the (compute requirements of analysis) entirely depends on the size of input. On larger strings it takes considerably longer and has rendered the browser completely unresponsive on low end devices.
Also, this is far from complete and the pre trained model is fairly inaccurate. I am mainly looking for collaborators to train this better.