semantic-web

This repo serves a pipelined tutorial that build up a knowledge base from web-crawled text using Protege and Apache Jena.

Prerequisite

1. Protege
2. Apache Jena
3. Apache Jena Fuseki

Check this link for installing Apache Jena & Fuseki (in Korean).

Installation

0. Install python dependencies.

pip install cython
pip install -r requirements.txt

1. Web scrap. (results in /data/raw)

python scrap.py

2. Explore scrapped data. (Results in ./data/stat)

python stat.py

3. Translate into Korean. (Optional. Results in ./data/translated)

python translate.py

4. Annotate the dataset. Insert any class label of your interest as a text span wrapped with doulbe squared bracket, e.g. [[Fruit]]. Also, correct the chunking of paragraphs, incorrect newlines charaters, and etc.

mkdir ./data/anno
cp -r ./data/raw/* ./data/anno
* Annotation with data/anno

5. Create a neat exel file for annotation. (Optional)

python anno2xls.py

6. Create OWL object classes & properties (T-Box) with Protege and place it under ./data/ontology. You can use a WebProtege to collaborate with your team.

   • ./data/ontology/root-ontology.owl

7. Populate individuls. (Results in ./data/ontology/basic.txt. Copy and paste the generated text to root-ontology.owl.)

python populate.py

8. Save root-ontology.owl as a turtle syntax.

   • ./data/ontology/root-ontology.ttl

9. Run Apache Jena Fuseki and write some SPARQL query you need. (Results in sparql/competency_questions)