Manual Pattern Entry for Title-based filtering

idea
#1

Currently it’s possible to select specific runs of words in the titles and add a like or dislike based on that run of words. It’s also possible to re-enter after saving this training and select another run of words to add a separate like or dislike about.
When I have cases where the title is “Why XXXX XXX XXXX is only the …” that I want to filter out of a feed, I would like to be able to manually add a regex filter to match this pattern. Currently I’d have to add a filter for all possible values of “XXXX XXX XXXX” since I need a unique contiguous run of characters for the filter. In this case a pattern like “^Why .* is only the” would be sufficient for performing the filtering I’m looking for.

From what I’m able to gather, pattern matches like this are already being run in the background for the Intelligence. I’d simply like to be able to add my own by manually typing it in sometimes, rather than only being able to have them auto-generated (which is incredibly handy on its own).

3 Likes
#2

I wish the same thing a lot

#3

I actuelly never used the training feature because I have no idea how it works and what impact is has.

I’d prefer a blacklist mechanism (with or without RegEx) and probably a whitelist mechanism optionally for highlighting.

Ultra-cool would be a scoring feature just like in advanced newsgroup clients back in the good old days: associate positive (upranking) and negative (downranking) integers to words and define a threshold for highlighting and hiding entries acccording to the sum of scores per item. Probably not for Jow Average but awesome control for tech-savvy users like me.

Edit: actually, I liked the idea of having a scoring mechanism as well so I wrote Scoring articles for highlighting and hiding