3

I'm looking for a decently sized dataset (more than 100k docs) which consists of bag of words documents along with a list of users where associated to each user is a list of those documents the user has in his or her library. Or even better, where associated to each user is a list of ratings that the user gave to those documents they viewed.

Does anyone know of such a dataset?

I'd really like to get my hands on the Mendeley DataTEL dataset, but they won't respond to my emails.

  • 1
    explain "bag of words" a little more please – albert Jul 26 '16 at 22:50
  • @albert, that just means the words in the documents don't have to be ordered, each document can just be a list of unordered words with their associated counts. Of course if they're ordered that's fine too, but they don't have to be. – Thoth Jul 27 '16 at 2:19
  • ....link to said mendeley datatel dataset? – albert Jul 30 '16 at 0:05
  • @albert, well I can't link to it, because it isn't openly accessible, but here is the link to the site that describes it: teleurope.eu/pg/pages/view/50632 – Thoth Jul 30 '16 at 1:47
  • 1
    @albert I'm not sure what that is. Ya like I said I emailed them and asked for it but they didn't respond. – Thoth Jul 30 '16 at 8:15
3
+25

While you may not have access to the Mendeley DataTEL dataset, have you considered other open datasets?

Netflix has an open API that provides a very similar set of information that I think might fit the needs of your request. The main features being:

  • Access to data for 100k movie and tv episode titles on DVD (as well as - Netflix account access on a user's behalf) with ratings information

Through the above, you could hypothetically back-in to a user's "library" by creating a list key-value dictionary that stores the UID of the user to a list of their rated interactions with Netflix media.

  • The problem is I need actual documents with text. This isn't just about finding a recommendation dataset with a user-item matrix, I need both the user-item matrix and I need each item to be a document made up of terms. I suppose I could use the synopsis of each movie as the document, is that what you're suggesting? – Thoth Aug 1 '16 at 3:17
  • Could the terms possibly relate to the genre information of the media being consumed by the user? i.e. tags related to genre, duration, year released. I imagine that you could also connect this to the IMDb site in order to get actor information if you really wanted to be creative. – shadow_dev Aug 1 '16 at 3:20
  • No I need actual documents, not metadata, the best idea would be probably to use the summary for the movie on wikipedia or IMDB or w/e as the document. – Thoth Aug 1 '16 at 3:21
  • There you go! Not sure if this exceeds the requirement of unsorted text - or if that was just to be made a widen options - but sounds like that should do it for you! – shadow_dev Aug 1 '16 at 3:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.