Lars-H / kg-analysis

Knowledge Graph and Linked Data uniformity analysis with affiliation networks

kg-analysis

Knowledge Graph and Linked Data uniformity analysis with affiliation networks

---

Setup

• Install Python development headers

apt install python3.7-dev

• Install Python C++11 bindings

pip install pybind11

• Install project dependencies

pip install hdt rdflib networkx numpy scipy pandas tqdm importlib

• Clone this project

git clone https://github.com/davidweisscode/kg-analysis.git

• Download HDT serialized datasets, indexfiles and ontologies from dbpedia.org and rdfhdt.org

mkdir kg/ && cd kg/
wget -c http://fragments.dbpedia.org/hdt/dbpedia2016-04en.hdt
wget -c http://fragments.dbpedia.org/hdt/dbpedia2016-04en.hdt.index.v1-1
wget -c http://downloads.dbpedia.org/2016-04/dbpedia_2016-04.owl

---

Getting started

Configuration

Specify your runs with a custom configuration of classes, data size, and projection approach.

config = {
    "classes": ["Athlete", "Artist"],        # List of DBpedia class names to analyze
    "project_method": "intersect",           # Choose between 'dot', 'hop', 'intersect', or 'nx'
    "kg_source": "kg/dbpedia2016-04en.hdt",  # Relative path to .hdt serialized Knowledge Graph
    "kg_ontology": "kg/dbpedia.owl",         # Relative path to respective Knowledge Graph ontology
    "subject_limit": 0,                      # SPARQL subject limit for each subclass (0 for unlimited)
    "predicate_limit": 0,                    # SPARQL predicate limit for each subject (0 for unlimited)
}

DBpedia classes

Analyze classes and its subclasses from the DBpedia class mappings.

Main Steps

Run the following four scripts in sequence together with your run configuration

1. Building

   • Query your dataset and build a bipartite Knowledge Graph for each Superclass specified in your config file
   • Run python3 build_graph.py run_config.py to output an edgelist in out/Superclass.g.csv

2. Projecting

   • Project your bipartite graph into its two onemode representations
   • Run python3 project_graph.py run_config.py to output onemode edgelists in out/Superclass.t and out/Superclass.b

3. Computing

   • Compute a KNC (k-neighborhood-connectivity) plot based on onemode graphs
   • Run python3 compute_knc.py run_config.py to output a KNC list in out/Superclass.k.csv

4. Analyzing

   • Get properties of the KNC plots computed beforehand
   • Run python3 analyze_knc.py run_config.py to save properties in your run's result file out/_results_run_config.py

python3 build_graph.py run_config.py >> log.txt && python3 project_graph.py run_config.py >> log.txt && python3 compute_knc.py run_config.py >> log.txt && python3 analyze_knc.py run_config.py >> log.txt

Results

The results of your run_config.py runs are saved in out/_results_run_config.py