The zeroth prototyp of a GND cache.

Clone the repo:

$ git clone [email protected]:miku/gndzero.git
$ cd gndzero

Create a virtualenv:

$ mkvirtualenv gndzero
...

Install dependencies (you'll need some C compiler):

$ pip install -r requirements.txt

Copy config.sample.py to config.py and adjust the values:

# TEMPDIR and HOME must reside on the same device
TEMPDIR = '/tmp'
HOME = './data'

Now create the sqlite datebase. This will take a while (download, extract, insert) - about 20-30 minutes:

$ python gndzero.py SqliteDB --local-scheduler

Run the server (development):

$ until python server.py; do echo "Re: (err: $?)" >&2; sleep 0.5; done

Examples:

• http://localhost:5000/gnd/118514768
• http://localhost:5000/gnd/121608557
• http://localhost:5000/gnd/4000362-0

Compare:

• http://d-nb.info/gnd/118514768
• http://d-nb.info/gnd/121608557
• http://d-nb.info/gnd/4000362-0

Format the output:

$ curl -s "http://localhost:5000/gnd/4000362-0"|xmllint --format -

Run the (gunicorn) server (production):

$ gunicorn --workers 4 --bind 127.0.0.1:5000 server:app

The local sqlite3 database is about 12G in size and contains 10004751 rows.

The GNDs form a graph, if occurences of GNDs in the metadata are interpreted as outbound links, hence you can compute a pagerank to find some often referred to concepts. An example output can be found here: https://gist.github.com/miku/7427351

$ time python gndzero.py HumanReadablePageRank --local-scheduler

The pagerank command line program can be found here: https://github.com/miku/gopagerank, with credits due to Thomas Dimson. The overall preprocessing for this takes too long (almost a day), but this is only a prototype.