Description
This homework spans two topics we recently covered. In the first task, you will extract data
from a structured source (tabular data) and model it using the RDF Data Cube ontology. In
the second task, you will use the AmpliGraph python library to generate knowledge graph
embeddings.
Task 1: Structured Data (6pts)
In this task you will extract quantitative data from World Bank data source (worlddevelopment-indicators.xlsx) and model the data using the RDF Data Cube ontology. The
data holds population numbers and population gender distribution percentages of more than
200 countries from 1960 to 2017.
• The general specifications and some modeling examples using the RDF Data Cube
ontology can be found here: https://www.w3.org/TR/vocab-data-cube/
Task 1.1 (3pts)
There are two dimensions (time period, country) and three measures (population, female
population percentage and male population percentage) in the data. Create RDF properties
to represent them in the Data Cube ontology and write your extension to world_bank_ont.ttl.
• Hint: inherit from qb:DimensionProperty and qb:MeasureProperty.
• For the sake of simplicity, you can use http://dbpedia.org/resource/Person and
http://dbpedia.org/page/Percentage as units for the measures above.
Task 1.2 (3pts)
Convert your data into RDF triples using the RDF Data Cube ontology with your extension.
Your RDF triples should include a dataset entity (instance of qb:Dataset), and list of
observations. Rename the file to world_bank_data.ttl for submission.
• You can test your RDF triples by executing the following SPARQL query and verifying
with the result below.
Query:
PREFIX qb: <http://purl.org/linked-data/cube#>
PREFIX my_ns: <http://dsci558.org/myfakenamespace#>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT ?population ?female_population ?male_population
WHERE {
?obs a qb:Observation ;
my_ns:refPeriod <http://reference.data.gov.uk/id/gregorian-interval/2014-01-01T00:00:00/P3Y> ;
my_ns:refArea / rdfs:label “Canada” ;
my_ns:measure_population ?population ;
my_ns:measure_femalePopulation ?female_population ;
my_ns:measure_malePopulation ?male_population .
}
Result:
population female_population male_population
35535348 50.39765 49.60235
2
Task 2: Knowledge Graph Embeddings (4pts)
In this task, you will use the AmpliGraph library to generate KG embeddings using different
embedding models and the given dataset file GoT.csv. We provide a python notebook
(kge.ipynb), which contains code and instructions to accomplish this task.
Task 2.1 (3pts)
Run the code under task 2.1 in the notebook, for each type of the models: TransE, DistMult
and ComplEx. You need to perform the following for each:
2.1.x.1. Train the model.
2.1.x.2. Evaluate it, include the performance scores in your final report.
2.1.x.3. Generate a visualization of the embeddings using Tensorboard (PCA-reduced, two
components) using the code in the notebook. Include a screenshot of the plot in your report.
• It is safer to re-execute the whole process for each model type after performing a
complete Kernel restart between each one of the sessions)
Task 2.2 (1pts)
Describe and explain (in your report, in a few sentences) what may cause the differences
between the different scores you got in 2.1.x.2 between the different models (i.e., compare
between the scores in 2.1.1.2, 2.1.2.2, 2.1.3.2)
Notes:
• Once you finish accomplishing all the tasks, change the notebook’s filename to
Firstname_Lastname_hw07_kge.ipynb (you will submit it).
• In order to explore the embedding space (in Tensorboard) with labels (instead of
indexes), you should load the file metadata.tsv (using the ‘Load’ button under the
‘Data’ tab). The file is generated using the last command in the notebook in the
resulting directory.
• You can find additional info on AmpliGraph here: https://docs.ampligraph.org/
Submission Instructions
You must submit (via Blackboard) the following files/folders in a single .zip archive named
Firstname_Lastname_hw07.zip:
• Firstname_Lastname_hw07_report.pdf: pdf file with your relevant written answers
• world_bank_ont.ttl: your extension of the Data Cube ontology in task 1.1.
• world_bank_data.ttl: your RDF triples in task 1.2.
• Firstname_Lastname_hw07_kge.ipynb: The notebook to accomplish task 2
• Firstname_Lastname_hw07_kge.pdf: A printed version of the notebook. You can save
your notebook to pdf using ‘Print Preview’ or ‘Download it as PDF’ in ‘File menu’
