arXiv category subscriptions

[breic]

Bill wrote: 

> I've looked over the design doc: the feature that I would like to add is a 'subscription' feature that lets a user display multiple categories at once. I primarily read quant-ph but occasionally check in on cs.CC and cs.CR, so such a feature is appealing to me. Comparing scores across categories is mostly meaningless, but I hope the good papers would float up anyway.

One way to manage subscriptions would be to have, for each category, fine-grained control of what papers appear in your feed: 

 quant-ph: 

  Subscribe X Only recommended -

  Show cross-lists X Only recommended -

  Show updates X Only recommended X

(Here, "only recommended" would only display papers that have received a favorite/scite; this could potentially be changed if we added a real recommendations engine.)  

One "cute" way of handling subscriptions would be with a preference box like: 

quant-ph Primary X Secondary X None X

cs.CC Primary X Secondary X None X

Primary = see all papers and replacements that someone has recommended/scited (cross-lists & replacements @ bottom)

Secondary = see only recommended papers (incl. cross-lists)

You should always see replaced papers that you faved earlier.  (The old Scirate didn't do this, but it would be very useful!)  

[jono]

I've made the following suggestion in the design doc, which I think covers some of this:

1. tagging (viewable, and inputable in all views and tabs).  del.icio.us stylie?
2. feeds
1. a feed can be an arxiv, a subcategory, a user, a tag, the entire arxiv
2. ability to add/delete a tab
3. ability to add/subtract a feed to each tab (e.g. quant-ph + CS.IT, but not stuff Aram likes, and with stuff Albert Einstein likes)
4. ability to make one of the tabs the default and otherwise re-order tabs
5. this could be implemented by drag/drop
6. perhaps one can add a threshold number of scilikes to each feed, with the most useful ones probably being 0 and 1.
   

[breic]

That sounds a bit overwhelming to me.  Maybe some of it can be built up over time.  If you look at, say, Facebook, the interface is now complicated, but it didn't start out that way.  There has to be a learning curve.  

Regarding tagging, I'm curious whether people think it would really be useful.  What is the incentive for tagging?  If you get into that, then you are competing with paper organization tools (Mendeley, Papers, Bibdesk, probably a half dozen others).  Delicious is dead for a reason. :)  

You can argue that these are just features, and that users don't have to use them if they don't want to, but each extra feature adds a cognitive burden that makes the site less compelling to use.  I feel the same way about rating papers.  Facebook, Google+, etc. all have binary "likes," not precise ratings systems, because the precision isn't worth the cost.  

[aram harrow]

I agree with a lot of this sentiment, and in particular I'm skeptical
of asking people to assign 1-5 stars because I think it may induce
paralysis.

However, I think tagging might be able to avoid this. An interface
that I think could be reasonable is something similar to the
functionality of the amazon.com wishlists: after you click "scite"
then something appears below giving you the option to apply one of
your existing tags, or type a new one in a text box. The text box
could then suggest common tags (delicious-style) as you start typing.
The incentive for a user is to organize their own reading into categories.


I agree that this is not necessarily the right way to go. Maybe it's
best to create a "reading list" feature (again like amazon's "list of
books recommended by X on topic Y") instead.

aram