Using intentional data becomes routine

I notice that IMDB has had a facelift. We use it a lot to see what folks think of movies - before renting or borrowing DVDs. We have a rule - rarely enforced ;-) - that we only watch movies with a score of 7 or over.

A part of the overhaul is the introduction of recommendations - "if you enjoyed this title, our database also recommends'. It is not clear to me upon what data the recommendations depend ...

I tried it with a few titles to see what the results were like. Worth having, I thought, although it is a little hidden on the end of the page.

I wonder have we now arrived at a stage where people expect this type of additional hint?

Related entries:

Update: Davey Pattern picks up this note and observes some other features including a keyword in cloud feature (see London). His note to this blog did prompt an interesting thought: good recommendations are good; other recommendations are sometimes interesting, and probably better than nothing. As part of the rich texture of suggestion we are becoming accustomed to.

Comments: 2

Mar 01, 2007
Dave Pattern

I feel so old this morning -- I can remember when IMDB was hosted on the servers at Cardiff University!


There's an option to add recommendations. Some of the recommendations seem a little random and their algorithm doesn't seem to take into account the user ratings given to films:


Psycho (1960) recommends the 1998 Gus Van Sant remake (rated 4.7) -- that's on a par with LibraryThing's UnSuggester in my opinion!


Vertigo (1958) recommends LotR: Return of the King -- a fine film, but one that has little in common with Vertigo?


Also, not sure how long this has been there, but clicking on a genre keyword now brings up a tag cloud - e.g. librarian. Clicking on a keyword in the cloud acts as a facet, so you can quickly track down all of those films featuring librarian nudity!

Mar 02, 2007
Bruce Newell

I think recommendations are now and have always been valued. I think most people at least look at recommendations (in Amazon, for instance) before buying. These recommendations may be ignored or followed, but the point is, I think most folks look at them.


Wouldn't it be wonderful to be able to harvest the borrowing patterns of the world's library users (in aggregate) to generate "If you liked this book/recording/film, you may like these others..."? The sample size would be large enough that the auto-generated recommendations lists would be, or rather, could be, spang on.