4 ms·
I’ve given up on content-based recommendation. The issue is that people’s interests in their topics and subtopics is primarily dictated by their individual cir
by ethn 6y ago
I’ve given up on content-based recommendation. The issue is that people’s interests in their topics and subtopics is primarily dictated by their individual circumstance and environmental real-world context. That is, if you are able to perfectly know the interests of an individual today, perfectly label the topic of all content, you’ll still give them incorrect content after a few days—-and with news article recommendations, by the next session of platform engagement (since they’ve sufficiently extracted the maximum value from the present topics by the nature of the session).
- nassimBreeze 6y agoMy idea allows for change of categories pretty much anytime. This user control maybe be able to circumvent the problem you're describing.
- ethn 6y agoThey’ll still have to sift through tons of no longer context-relevant content. You’re also expecting the user to rationally know his context in terms of your chosen categories, and worse, to put in the work to formalize them for your engine at every session. There is no information in any content or a previous interest matrix which guarantees any real-world relevance in the next session. Instead, it’s likely the user exhausted the utility of their previous topic matrix. The effect is that the moment you determine content relevant to their context, they’ve already been sufficiently informed. I’ve tried algorithms far more advanced with ML/Lebesgue measure theory, which were technically optimal, but with practically laughable results.
- jacobobryant 6y agoI'd be interested to hear your thoughts on this: https://findka.com/blog/essays/ https://findka.com/blog/essays/ (a project idea I'm thinking about doing). I wonder if preference for essays/evergreen content (as opposed to news articles) would be less affected by day-to-day context.