Train based on text in body

Apart from Title, Author, and Site, would it be possible to add a param to support training based on specific text in the article body.

Example:
For this feed (https://newsblur.com/site/7714020/reverbmarkets-youtube-videos), I’d like it if we can train based on ‘Andy Martin’ (his demoes are dope!).

Thanks in advance for your consideration!

2 Likes

@samuelclay I see you’ve mentioned that this will be computationally expensive for all articles.
How about at least for YT videos/description fields.
I’m sure they’re a lot more manageable than whole articles.
Bulk of the feeds are more text based rather than YT feeds.

Thoughts?

Am I the unlucky one?
All new posts are blessed by a response except mine :frowning:

I’m back here to ask for better filtering again. You’re not alone, this has just been asked already. I actually came back here to be more specific, and ask that the Newsblur Intelligence trainer let us choose which feeds we want to train, or at least to be able to do them all, the ones I need to filter most never showed up in the intelligence trainer. The Verge, Ars Technica, those feeds are blowing up right now due to Apple’s Time Flies event.

you can train specific feeds. open the feed and the drop down the menu at the feed name in the upper left.

image

you can also hit t with any story open to train specific attributes of that story for the feed it’s in

Ah ok, oops, don’t know how I missed that. I can’t train by adding keywords, it only lets you train by author and whatever tags it picks up. It also still matches too easily.“Sea” matches “Sean McConery” for example.

I do see keywords I have trained in the past, but I’m noting that I can’t delete them. Because of how the keywords match far more than desired, I would actually like to delete some keywords as well as add some others manually without having to pull up a specific story with a keyword I’m looking for.

I would love to be able to train the filter/intelligence at will a lot more than I currently am able to. For example with the Apple Time Flies event I had like over 100 tech related emails, and I’d love to be able to still get some content from The Verge or ARS Technica on other topics besides Apple stuff since I’m already sated on the topic. I wish I could set a filter and give it an expiration date at which point the filter would just disappear.

Pardon me for hijacking this thread Op. I like your idea though. It’d be great if we could train in an allow mode or block mode. Block mode = nothing gets through unless it has a desired parameter or keyword for the RSS feed. It’d be such a drag to have to thumbs down all the authors for some RSS Feeds that have a lot of contributors. I’m not even sure how YouTube feeds work, would you even get author tags in those feeds? How do you even do that, do you copy the subscribe URL or something for a YouTube channel?

you can select parts of the title.

toggle through thumbs up, thumbs down, and neutral (delete).

So now that highlighting text has launched, I finally have the interface I need to be able to ship this feature. It’ll be premium only of course, but it shouldn’t be a problem to implement. I’ll prioritize it now and see if I can get to it this winter when I’m back on feature work.

Here’s the ticket:

3 Likes

Looking forward to further feature development. The rest of this message is from last night, couldn’t post as I got booted from the 'puter.

Oh I see, once a user marks a keyword as neutral and saves, it gets deleted. That’s a bit counter-intuitive, but good to know.

I was thinking, It would be so useful if the author tag could be used to see the articles that the author has attributed to them, I could then see if I like their work, and then decide to focus their articles, filter them out of focus mode, or leave it as neutral.

More ideas for consideration. Thanks for a great product, keep it up :).

Love this idea. We also have the story list header bar for it now that we have Saved Searches. Not sure how to launch you into seeing a single author-view though. Tapping on the author will train it.