Sebastian Neumaier, Jürgen Umbrich, Josiane Xavier Parreira, and Axel Polleres. In Proceedings of the 15th International Semantic Web Conference (ISWC 2016), Kobe, Japan, October 2016. [ pdf ]
We apply a hierarchical clustering over information taken from DBpedia to build a background knowledge graph of possible “semantic contexts” for bags of numerical values, over which we perform a nearest neighbour search to rank the most likely candidates.
The total setup-time for all 50 properties in props.csv takes 15-30 minutes and ~20GB of RAM. In order to test the system without this extreme built time and requirements use only a small subset of properties with a lower number of corresponding subjects.
$ git clone https://github.com/sebneu/number_labelling.git$ cd number_labelling- (optionally) setup virtual environment
$ virtualenv --system-site-packages labelling_env$ . labelling_env/bin/activate- Install anycsv CSV parser
$ pip install git+git://github.com/sebneu/anycsv.git- Install requirements
$ python setup.py install- Setup local files
$ tar -xzf local/common_types.tar.gz -C local$ cat local/subjects.tar.gz.* | tar xzvf - -C local- Run API service
$ ./runner -hto show help$ ./runner -c config.yamlto start the API service- Example curl request:
$ curl -X POST -F csv=@testfile/stadiums.csv http://localhost:8081/labelling?column=2