sp1ff
FEATURE REQUEST: Add adaptive scoring
Current scoring is OK as a starter but what about adding some adaptive scoring ?
As a user, I'd like to automatically adjust scoring based on simple rules:
- if pressing r in the search buffer (aka, not reading the entry) should decrease the score for all words of title by, say -1, and also the author by -1
- if pressing k (kill ?) decrease by 5
- viewing the entry (RET on the entry) should increase by 1 (for both words title and author)
Etc.
WDYT ?
For inspiration, see this: https://cce.whatthefuck.computer/elfeed-adaptive-scoring
Yeah, that's been in the back of my mind for some time. The inputs seem simple: as you've already noted, marking as read or killing indicate a lack of interest, reading an entry in elfeed, or opening the entry in a browser indicates interest. The question, however, is what to do with that information? If the user has indicated a lack of interest in a particular post, the trick is understanding why; what aspect of that post should be recorded and matched against future posts?
Ryan's post (love his site, btw) uses "large" words in the post title, but that's just one choice among many. I was thinking more in terms of a general-purpose classifier, which would run locally, on training data stored locally (i.e. elfeed-score shouldn't be "a fucking weirdo about it"), but I haven't taken the time to think about the features to be extracted from posts, yet.
Any suggestions?
