search

cedricbonhomme / newspipe

A web news aggregator.
https://www.newspipe.org
GNU Affero General Public License v3.0
437 stars 41 forks source link

Article-level categorisation? #32

Closed ghost closed 8 years ago

ghost commented 8 years ago

I just started playing with newspipe.org, nice service overall. I'll keep playing, for now. But, I have noticed a bad abstraction; feeds are categorised, instead of posts. Also, "drop filters" are missing.

For example, what if I want to subscribe to 10 major newspapers, but I only care about posts that mention "bitcoin" or "privacy", or alternatively "biotech" or "diybio" (two areas I have an interest in). I'd like to be able to define filters like:

  1. If (regex '(blockchain|privacy)' then categorise->privacy
  2. else if (regex '(biotech|diybio)') then categorise->biotech
  3. else delete post (drop)

This is because many feeds are totally heterogenous, and you only care about a subset, or several subsets. And with RSS the problem isn't getting information, it's filtering it.

Some feeds are homogenous in topic, and for those there should still be a "feed category", but I feel that under-the-hood this should just mean a default filter for all posts, and that filters should still operate on posts, not feeds.

Thanks!

cedricbonhomme commented 8 years ago

In order to stay more coherent with the OPML specifications I will keep the categorization of feeds and not articles (posts). Making a kind of abstraction wasn't my goal here. It is just a way to sort the feeds, like its done with a lot of newsreaders. In the OPML export of Newspipe, articles are not included.

Filters are peer feed, but executed on the articles.

If in those 10 majors newspapers I only want articles about 'blockchain' I'll make a filter which will mark as read articles that do not contains the word 'blockchain'. Or maybe, mark articles containing 'blockchain' as 'to_read' (not yet implemented). For the moment the filtering is done on the content of the article. Maybe I'll have to also take into account the tags associated the articles.

Anyway, I like the idea which consist of creating filters that are applicable for all feeds.

ghost commented 8 years ago

Cool, thanks for the feedback. It's still something I'd love to see, but I understand why it conflicts with the goals of the project.

Looking forward to seeing where you take newspipe! :)

cedricbonhomme commented 8 years ago

You're welcome. As I said I found interesting the idea to have filters applicable for all posts (of all feeds).

A thing that I already have tested is k-means clustering of articles. This gives a kind of automated clusterization of articles. A possible direction...

ghost commented 8 years ago

Alrighty; seems your design goals are clear on the matter, and it's only as suggestion. :)

Nice work on newspipe, looking forward to seeing where you take it!

closes #32