I'd appreciate some feedback on how Amphetarate fits your feed reading process. Is AmphetaRate useful for you? Was it something else that you expected from the service?
Thanks,
Kyvinh
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I started using AmphetaRate in RSSOwl today. I've been playing around quite a bit and rated several posts.
I like the basic idea as it is a neat way to find new content or even feeds. Generating a custom feed that suits my preferences sounds great.
[I feel like the following comments belong here rather than on RSSOwl]
1/ I keep on seing posts in my "Recommended Article" feed that I've rated before (and therefore read). This is really annoying. There has to be a way to exclude myself from the list of nearest neighbours, right?.
Sometimes even posts I rated "bad" are marked as "good" (some other users must have liked them). How is this rating stored? Or which information is stored, because there is no global rating of a post if I understand correctly.
2/ "Latest best news" essentially shows my ratings, which might be related to the rather small number of subscribers (about 700) and my heavy rating activity. The number of rated news items on the site is not updated in real time. Why is that? (It is not due to local caching...)
3/ Obviously, it requires training with several messages to produce good recommendations. This might be a problem for new users. I will continue to use it and see if the results get more specific over time.
Dave
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
For 1/, that should not happen. The system should really not recommend you articles you've rated. At this point, I can only assume that the rating has not been sent correctly because it never happened before. To check if your rating has indeed been sent, check if any of your ratings appearing in "Latest best news" also appear in your recommendation feed. That will tell us if the rating has been sent correctly. Also, send me your user ID and some of the problematic news titles at kyv@email.com and I will look into it.
2/ Yes, "Latest best news" is not as useful as can be, because of the low number of active users as you said (although I am confident that the latest changes to the algorithm will bring more users). The news displayed on the website is updated hourly, for page loading speed.
3/ Indeed, it requires training, by definition of "personalization". Recently, the training has been minimized through the use of a bayesian recommender... So thanks for trying it out.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm in the first few days of training, and have noticed quite a lot of suggested / training items in a language I don't read. I've been giving these the lowest rating, assuming that'll do the trick. I don't think I've done enough just yet, but in general a language preference might be a good idea so that items in another language don't need to be trained out.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'd appreciate some feedback on how Amphetarate fits your feed reading process. Is AmphetaRate useful for you? Was it something else that you expected from the service?
Thanks,
Kyvinh
I started using AmphetaRate in RSSOwl today. I've been playing around quite a bit and rated several posts.
I like the basic idea as it is a neat way to find new content or even feeds. Generating a custom feed that suits my preferences sounds great.
[I feel like the following comments belong here rather than on RSSOwl]
1/ I keep on seing posts in my "Recommended Article" feed that I've rated before (and therefore read). This is really annoying. There has to be a way to exclude myself from the list of nearest neighbours, right?.
Sometimes even posts I rated "bad" are marked as "good" (some other users must have liked them). How is this rating stored? Or which information is stored, because there is no global rating of a post if I understand correctly.
2/ "Latest best news" essentially shows my ratings, which might be related to the rather small number of subscribers (about 700) and my heavy rating activity. The number of rated news items on the site is not updated in real time. Why is that? (It is not due to local caching...)
3/ Obviously, it requires training with several messages to produce good recommendations. This might be a problem for new users. I will continue to use it and see if the results get more specific over time.
Dave
Hi Dave,
Yes, your comments belong here.
For 1/, that should not happen. The system should really not recommend you articles you've rated. At this point, I can only assume that the rating has not been sent correctly because it never happened before. To check if your rating has indeed been sent, check if any of your ratings appearing in "Latest best news" also appear in your recommendation feed. That will tell us if the rating has been sent correctly. Also, send me your user ID and some of the problematic news titles at kyv@email.com and I will look into it.
2/ Yes, "Latest best news" is not as useful as can be, because of the low number of active users as you said (although I am confident that the latest changes to the algorithm will bring more users). The news displayed on the website is updated hourly, for page loading speed.
3/ Indeed, it requires training, by definition of "personalization". Recently, the training has been minimized through the use of a bayesian recommender... So thanks for trying it out.
How many posts does it typically take before posts stop showing [TRAINING]? My ID is TYY879
I'm in the first few days of training, and have noticed quite a lot of suggested / training items in a language I don't read. I've been giving these the lowest rating, assuming that'll do the trick. I don't think I've done enough just yet, but in general a language preference might be a good idea so that items in another language don't need to be trained out.