This project is a first attempt to empower you to have your own homebrew movie repository that you can use most effectively. The iMBD's dataset is incredibly complete, but it's API can be a pain to use, and the data formats they use to distribute their data are (literally) from decades ago. There is also tMDB, that you can also fetch for free, but according to their lengthy user license, you can't even keep a cached copy in your machine.

We have a lot of information on DBPedia, FreeBase and LinkedMDB, but each one of these services uses a different vocabulary, and it's very hard to integrate information from all this data.

Enter Hombrew: the Homebrew Movie Repo. With it, you can create a JSON-LD representation of existing movie data sources (for now, IMDb and Freebase are supported), and integrate them right away with whatever it is you need. If you like the idea, you too can create a JSON-LD parser for your favourite movie source using the schema.org vocabulary, and submit a pull request.

First, make sure you have Python 2.7 installed with all the packages in requireents.txt installed to your current environment.

1. Create a directory to hold your iMBD stuff, and download the imdbpy2sql.py script from Github in it. Start downloading the iMDB files, it's about 1.5GB large.

   $ pip install -r requirements.txt $ mkdir temp && cd temp $ wget https://raw.githubusercontent.com/alberanid/imdbpy/master/bin/imdbpy2sql.py $ wget -r --accept="*.gz" --no-directories --no-host-directories --level 1 ftp://ftp.fu-berlin.de/pub/misc/movies/database/

2. Once you have all the files, generate the sqlite database. That should take about 45 minutes using an SSD disk. Make sure that the imdbpy package is installed in your environment.

   $ python imdbpy2sql.py -u sqlite:./imdb.sqlite -d temp.gz --sqlite-transactions

3. With your fresh imdb.sqlite, get some indexes in it, because this database is quite large and we'll query these fields all the time.

   $ sqlite3 imdb.sqlite sqlite> CREATE INDEX title_idx_kind_id on title (kind_id); sqlite> CREATE INDEX movie_info_idx_itid on movie_info_idx (movie_id, info_type_id);

4. You're all set! Check the parser_imdb.py script to make sure it can find your imdb.sqlite, and run the script. It will output the most popular movies in the JSON-LD format. You can pipe the script to a file and let it run overnight (because boy, this takes a while! 😴 😩 😡 )

Run the script and you'll see it grab triples from LinkedMDB right away from the internet, requesting a conversion to the schema.org vocabulary on the fly. (Vive le Open Data 😎 ). The problem is, for not, it looks like their sparql endpoint won't manage to let us get a dump all at once without crashing.

Most importantly, we have to fix the LinkedMDB script to get data in bulk, and then integrate the dataset from the two services, possibly using MongoDB to keep things simple and fast. Later on the sky is the limit, we can create parsers for every dataset out there.

© 2015/07 Wladston Ferreira Filho under the BSD License.