# Would appreciate more robust Hide and Focus **URL:** \ **Category:** problem\ **Created:** [April 8, 2020, 10:30pm UTC](https://forum.newsblur.com/t/would-appreciate-more-robust-hide-and-focus/8095 "2020-04-08T22:30:45Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![Adam\_Logan](https://forum.newsblur.com/user_avatar/forum.newsblur.com/adam_logan/32/255_2.png) [@Adam\_Logan](https://forum.newsblur.com/u/Adam_Logan)\ **Post date:** [April 8, 2020, 10:30pm UTC](https://forum.newsblur.com/t/would-appreciate-more-robust-hide-and-focus/8095/1 "2020-04-08T22:30:45Z")
I’ve been using the Focus feature a lot lately, mostly to filter out terms such as “Covid-19” “Coronavirus” “Pandemic” as there is way too much volume of such news. Asides from hiding content, I’ve also been working on training with terms I want to see more of. The behavior I’ve noticed is that matching is not intelligent. For example I added a positive filter (want to see more of) with term “sea”. I saw an article pop up for Diseases, um, no, that’s not what I want… So my request is for a more robust training be it Hiding or Focusing. I have a beginner’s level of competency with Regex, it’d be neat to be able to use Regex query strings to train Article titles, although I understand that this would probably be excessive, and probably resource intensive. At the very least, I’d like to see better baseline behavior though, the default search query should include word boundaries. I think case matching would be inadvisable though. ![2020040719092406](https://forum.newsblur.com/uploads/default/original/1X/eea24599bb8d59e0c2df7f41c2fe6fcf3ad4ca22.jpeg) ---
**Author:** ![samuelclay](https://forum.newsblur.com/user_avatar/forum.newsblur.com/samuelclay/32/857_2.png) [@samuelclay](https://forum.newsblur.com/u/samuelclay)\ **Post date:** [April 9, 2020, 2:47pm UTC](https://forum.newsblur.com/t/would-appreciate-more-robust-hide-and-focus/8095/2 "2020-04-09T14:47:39Z")
I agree, word boundaries in regex would be a great change from the current system of string matching. These are tricky to launch because the intelligence part is distributed to every client, so it has to change in four places at the same time, which is why I resist any changes. I’ve added it to the list and if I ever get around to hiring a backend developer this might actually happen: > > > See https://forum.newsblur.com/t/would-appreciate-more-robust-hide-and-focus/809…5. > > Basically move the title intelligence classifier from string matching to a regex. Would have to be done simultaneously on the backend, web, iOS, and android. Every native client has training in app so they need to know how to calculate unreads. Maybe we can rip that out and add an endpoint called every time a classifier is trained for the new unread status of all visible stories. ---
**Author:** ![Adam\_Logan](https://forum.newsblur.com/user_avatar/forum.newsblur.com/adam_logan/32/255_2.png) [@Adam\_Logan](https://forum.newsblur.com/u/Adam_Logan)\ **Post date:** [April 9, 2020, 3:54pm UTC](https://forum.newsblur.com/t/would-appreciate-more-robust-hide-and-focus/8095/3 "2020-04-09T15:54:53Z")
Thanks 👍