What recommendation datasets, that are available, are considered
"large" by Mahout testing standards? Yahoo KDD Cup is offline, the
Netflix data went under a cloud...
--
Lance Norskog
goks...@gmail.com
Those are both reasonably large, but not commercial in scale.
At Veoh, we had about 10 non-zero elements in our raw data. I think Netflix
has 100 million.
On Thu, Jul 7, 2011 at 8:05 PM, Lance Norskog wrote:
> What recommendation datasets, that are available, are considered
> "large" by Mahout
There is still a libimseti dataset
http://www.occamslab.com/petricek/datawith 17,359,346 ratings. People
are scared after the Netflix lawsuit.
On Thu, Jul 7, 2011 at 10:17 PM, Ted Dunning wrote:
> Those are both reasonably large, but not commercial in scale.
>
> At Veoh, we had about 10 non-zer
Is it taken offline as well ?
On Thu, Jul 7, 2011 at 10:40 PM, Alex Kozlov wrote:
> There is still a libimseti dataset
> http://www.occamslab.com/petricek/datawith 17,359,346 ratings. People
> are scared after the Netflix lawsuit.
>
> On Thu, Jul 7, 2011 at 10:17 PM, Ted Dunning
> wrote:
>
> >
The link is http://www.occamslab.com/petricek/data/
The KDD or Netflix data are plenty big to play with. How big is big for your
purpose?
On Fri, Jul 8, 2011 at 7:05 AM, web service wrote:
> Is it taken offline as well ?
>
> On Thu, Jul 7, 2011 at 10:40 PM, Alex Kozlov wrote:
>
> > There is st
Another dataset to play with is this compilation of song listenings
scraped from the last.fm API:
http://mtg.upf.edu/node/1671.
Should include about 20M ratings.
--sebastian
On 08.07.2011 09:17, Sean Owen wrote:
The link is http://www.occamslab.com/petricek/data/
The KDD or Netflix data are
Thanks.
Netflix & Yahoo KDD were my first choice, but are gone. It did not
occur to me that stashing such things away would be wise; packrat
though I am.
Purpose is testing large user/item or document'/term databases.
On Fri, Jul 8, 2011 at 12:44 AM, Sebastian Schelter wrote:
> Another dataset
Movielens would be the one thats most commonly used by researchers, they
have a 100k, 1million and 10 million ratings dataset.
On Fri, Jul 8, 2011 at 10:26 AM, Lance Norskog wrote:
> Thanks.
>
> Netflix & Yahoo KDD were my first choice, but are gone. It did not
> occur to me that stashing such
It's not a traditional ratings corpus, but the ASF mail archives I put up all
have clear provenance and are freely available and I don't think it is too hard
to make a recommender problem out of them, likely based on the replies. There
are 6m+ items in it. And now that Amazon has free inbound,
Ratings and more generally "parallel universe" or "dual space" or
"dyadic" (but that is other things): Correspondences between samples
in two different parallel spaces.
A mail corpus has different kinds of gleanable knowledge: word/subject
line correspondences, authoritative mail v.s. conversation
On 9 July 2011 00:03, Lance Norskog wrote:
> Ratings and more generally "parallel universe" or "dual space" or
> "dyadic" (but that is other things): Correspondences between samples
> in two different parallel spaces.
>
> A mail corpus has different kinds of gleanable knowledge: word/subject
> lin
Thanks. That final one is now in torrent-land.
The RDF SKOS stuff reminds me: RDF seems pointless without a
confidence factor per triple. It would be cool to take a large
self-inconsistent RDF-triple graph and grind out a set of confidence
factors that makes it consistent. "This stuff is good, thi
There's a Mendeley data set available on request:
http://dev.mendeley.com/datachallenge/
5M documents (scientific papers etc.) x 50K readers -- not huge, but
an interesting problem space.
On 8 July 2011 23:23, Dan Brickley wrote:
> On 9 July 2011 00:03, Lance Norskog wrote:
>> Ratings and more
13 matches
Mail list logo