Show HN: Pick a movie you like and discover others semantically | Hacker News Reader

Show HN: Pick a movie you like and discover others semantically

movies.ootu.com

8 pointsmhuffman10 comments

This is cool. Movie recommendations are not easy, since there are just too many dimensions with which to compare movies. Sometimes there is just no better movie recommendation than just having a friend who knows your tastes.

Here is another site that has proven some worthy suggestions for me in the past. http://www.nanocrowd.com/

We have a broader semantic navigation technology that is driving this site. Shortly we will be doing apps for books, video games, and business.

We think that why movie recommendations are so hard is because people might like the same movie for different reasons. For example, if I told two friends that I liked the movie "Titanic", and asked for a recommendation they might separately recommend "The Perfect Storm" and "Avatar". Two very different movies, but both would count as good recommendations for different reasons. The problem, is that when a person recommends a movie to you, they have already filtered it according to what they thought was interesting about the movie.

Having said that it is still the best way to discover new movies. We believe that easy to use semantic browsing is a good middle-ground for any information discovery.

nanocrowd is very cool. I have not heard of the company before, but they have a very interesting take.

What's the source dataset? IMDb?

No we are using open semantic data sources to start. Soon movie lovers will be able to add their own data to the ontology.

Open semantic data sources? That's very vague. :) Freebase?

Freebase, wikipedia, yago, wordnet, and some others. Nearly all hand-curated.

YAGO looks interesting -- and it seems to encapsulate both Freebase and Wikipedia. Does this mean you're using YAGO to get to Freebase/Wikipedia?

No. YAGO is actually a curated subset of those.

Do you'll do TV shows too?

Right now just movies, but if there is interest TV shows would not be a problem. The underlying technology can work with most semantically related data sources.

Reply on news.ycombinator.com