Ability to train a feed by keywords selected by NewsBlur

Right now you can select keywords from the title, which is good, and from those specified in the article, which is also good. However, many feeds don’t have keywords specified in the article and the title isn’t specific enough to train against. What would be helpful for me would be for NewsBlur to scan the article for keywords – there are a variety of algorithms out there for this – and present them for training in the same interface as article-specified keywords. I know it wouldn’t be perfect, but it would probably be 95% of the way there.

Sorry if this is a duplicate idea. There are a lot of keyword ideas in the system and I didn’t read them all.

1 Like

The reason I don’t do this right now is that the story content is stored in a compressed format, which takes a while to go through during story unread counting. I also wonder how much benefit there is to finding keywords in story bodies, since so few people would use it but it would take an enormous amount of developer attention.