ECS7028U Data Semantics Final Coursework: Semantic data modelling and Ontology design
ECS7028U/P: Data Semantics – Mini Project
(Final Coursework)
This coursework makes up 60% of the total marks for the module.
The objective of this min-project is to put your semantic data modelling, ontology design, logic, data validation and rule-based reasoning skills into practice that you have learnt in the lectures and labs.
The coursework is design to be open-ended, that is, you must make your own choices when it comes to the data you choose to work with and discover your own data sources (examples and possible data sources are provided on the QM+ module page) and choose the best way to document your work (report and code to be submitted).
Your task is to define, populate and query an ontology (including A-Box and a T-box) on a topic of your choice. The ontology must be able to integrate and reuse already available semantic data. At least two concepts of the ontology T-Box must be taken from external semantic data repositories. This way, the ontology will have an A-Box that can be populated with already existing data. You will use Protégé to design the ontology, and Python-based semantic tooling to populate the ontology with real world data.
You should achieve the following specifically (compulsory for full marks):
Core Task (basic data engineering): Define your ontology using OWL2. The T-Box must be created using Protégé and should be your own work (not an existing ontology but may import existing ontologies). Populate the knowledge base from an external semantic data repository using SPARQL 1.1. Verify that you can also query the local ontology using SPARQL.
Extra Task (data integration): Complete data gathering as above, but your ontology should fuse information from at least two distinct external data repositories. The query to your local ontology should answer questions that cannot be answered by either remote knowledge base alone.
Advanced Task (reasoning): You are required to use Description Logic (DL) rules to define as many concepts as possible with the help of SWRL rules to compensate for the limitations of the Protégé inference engines with DL. The A-Box (individuals) must be created in a way to demonstrate the correctness and effectiveness of the logic rules defined.
All tasks are compulsory for full marks.
You should submit a zip file with the following elements:
§ A report: PDF document describing how you constructed the ontology: you should say where you got the data from, and you should also say what difficulties you encountered and how you solved them. The document must have also a final section explaining what source code files and models are included, and the required steps to run the code. You are expected to submit a report explaining
your assignment. If you fail to submit a report with your ontology and python files, you might receive 0% as it is the only way of proving the work is yours.
§ A Protégé-OWL ontology (.owl file)
§ A python script (.py) that can be used to populate the ontology from a SPARQL
endpoint.
§ Another python script that queries the local store to demonstrate to the user that information can be easily accessed. To test the system, the user should be able to execute any arbitrary query supported by your ontology.
Marking criteria:
Core task (60% coursework marks): Correctly designed ontology with basic taxonomy and property hierarchy (15%), correct domain and range restrictions (5%), correct and effective use of object properties (including constraints and characteristics such as functional, transitive and irreflexive), correct and effective use of data properties, logical and correct use Description Logic to define concepts (10%), Ontology population python script and SPARQL queries (20%), justification explanation and validation of the ontological modeling decision in the report (10%).
Extra task (20% coursework marks): Use of appropriate, diverse data sources (5%), correct mechanism to retrieve and transform data to fit your ontology (10%), explanation of the mechanism in the report (5%). The data sources can be RDF dumps (e.g., loaded into a local database) and SPARQL end-points and may include one non-semantic dataset you convert into RDF locally for it to be queried.
Advanced task (20% coursework marks): Correctly working (inferencing) ontology with one of the reasoners provided with Protégé and use of SWRL rules: object and data properties and SWRL built-ins (10%). A correct A-Box with enough individuals to use with the defined logic rules, expecting majority of the relations defined by the data properties and concepts to be inferred by the engine, not hard coded (5%), correctly commented rules and explanations in the report (5%).
Fig. 1. Example SWRL rules in Protégé using the SWRLTab plugin
Further details and support:
This assignment is intended to be open ended and exploratory in nature. However, for illustration, some examples of possible tasks could be:
-
create and populate an ontology covering movies and cities which could be queried to find movies filmed in cities with a population less than 1M; or
-
create and populate an ontology about companies including location, employees, and profits, which could be queried to find the UK based companies with the largest profit per employee.
Some publicly available semantic web data sources that can be useful for this
coursework are listed below:
-
Dbpedia (https://www.dbpedia.org/) provides an RDF version of the information available in the regular Wikipedia. It also provides an SPARQL endpoint for remote access https://www.dbpedia.org/resources/sparql/
-
You can access governmental datasets from data.gov.uk and data.gov. In some cases, you might need to download the dataset, as they don’t provide a SPARQL endpoint.
-
Wikidata (https://www.wikidata.org/) is a free and open knowledge base that can be read and edited by both humans and machines. Wikidata acts as central storage for the structured data of its Wikimedia sister projects including Wikipedia, Wikivoyage, Wiktionary, Wikisource and others, with a SPARQL end point: https://query.wikidata.org/
-
Datasets with SPARQL end points: https://www.wikidata.org/wiki/Wikidata:Lists/SPARQL_endpoints
-
Over 10,000 arbitrary datasets from https://datahub.io/search (some are in RDF).
