Computational report writing
Budget: ₹600 – ₹1,500 INR
Computational Modelling in the Humanities and Social Sciences
The House of Commons SPARQL endpoint (https://api.parliament.uk/sparql/) provides access to
structured data about Members of Parliament (MPs) and questions that have been asked in the House
of Commons. Use this data together with data from Wikidata (where necessary) to answer the following
questions, presenting your findings in the form of a report of up to 3000 words, based on data about
questions asked by MPs between 1 January 2023 and 30 September 2023 inclusive:
Q1. To what extent do Members of Parliament (MPs) tend to ask questions that directly reference
their own constituency or a location in it? You should answer this question by identifying named
entities that refer to places or identifiable geographical features (e.g. “Dartford Crossing”,
“Reading Gaol”, etc.) in asked questions, and determining whether or not these are located in
the MP’s constituency using data from Wikidata. You are free to choose any reasonable method
in doing so – even if doing so will result in some false negatives – (e.g. relying on Wikidata
property P131 “located in the administrative territorial entity”), however you should estimate
how reliable you believe your chosen approach is. (N.B. it is not expected or required that your
approach will result in 100% accuracy.)
Q2. By applying LDA topic modeling and analyzing the results, what (if any) identifiable regional
differences are there in the types of questions asked – e.g. do MPs representing, say,
constituencies located in the North of England tend to ask more questions about certain topics
than those in Southeast England? In answering this question, you should start by aggregating
data into regions larger than an electoral district, such as those denoted by the property “region
of England” https://www.wikidata.org/wiki/Q48091 (For simplicity, you may treat Scotland and
Northern Ireland as two separate regions without further subdivisions, or alternatively use any
reasonable administrative subdivisions for these regions as you see fit). Discuss the assumptions
and limitations of your approach and analysis.
The following SPARQL query can be used as a starting point:
SELECT *
WHERE {
?question <https://id.parliament.uk/schema/writtenQuestionIndexingAndSearchUin> ?qnum .
?person <https://id.parliament.uk/schema/askingPersonHasQuestion> ?question .
?question <https://id.parliament.uk/schema/questionText> ?text .
?question <https://id.parliament.uk/schema/questionAskedAt> ?date .
FILTER (?date >= "2023-01-01+00:00"^^xsd:dateTime && ?date < "2023-10-01+00:00"^^xsd:dateTime)
}
This query returns the following data:
?question Entity representing a question
?person Entity representing the person who asked the question
?qnum Numerical identifier for this question
?text The text of the question
?date The date the question was asked
Your report must:
1. Document how data was collected, and briefly state any relevant modules or toolkits that were
used. If using SPARQL queries not contained in your code (other than the one provided), you
must include these as separate files within your code submission.
2. Describe your implementation and the models constructed.
3. Summarize your results in appropriate ways, including use of suitable visualizations.
4. Critically evaluate the models. This must involve critical analysis of the adequacy of the modelling
(e.g. what are the assumptions the models rely on, and do they all hold; are there factors or
biases that might invalidate conclusions drawn?), but should also involve some comparison with
external data and/or published social/political science research on relevant subjects.
5. Discuss what conclusions can be drawn from the
The House of Commons SPARQL endpoint (https://api.parliament.uk/sparql/) provides access to
structured data about Members of Parliament (MPs) and questions that have been asked in the House
of Commons. Use this data together with data from Wikidata (where necessary) to answer the following
questions, presenting your findings in the form of a report of up to 3000 words, based on data about
questions asked by MPs between 1 January 2023 and 30 September 2023 inclusive:
Q1. To what extent do Members of Parliament (MPs) tend to ask questions that directly reference
their own constituency or a location in it? You should answer this question by identifying named
entities that refer to places or identifiable geographical features (e.g. “Dartford Crossing”,
“Reading Gaol”, etc.) in asked questions, and determining whether or not these are located in
the MP’s constituency using data from Wikidata. You are free to choose any reasonable method
in doing so – even if doing so will result in some false negatives – (e.g. relying on Wikidata
property P131 “located in the administrative territorial entity”), however you should estimate
how reliable you believe your chosen approach is. (N.B. it is not expected or required that your
approach will result in 100% accuracy.)
Q2. By applying LDA topic modeling and analyzing the results, what (if any) identifiable regional
differences are there in the types of questions asked – e.g. do MPs representing, say,
constituencies located in the North of England tend to ask more questions about certain topics
than those in Southeast England? In answering this question, you should start by aggregating
data into regions larger than an electoral district, such as those denoted by the property “region
of England” https://www.wikidata.org/wiki/Q48091 (For simplicity, you may treat Scotland and
Northern Ireland as two separate regions without further subdivisions, or alternatively use any
reasonable administrative subdivisions for these regions as you see fit). Discuss the assumptions
and limitations of your approach and analysis.
The following SPARQL query can be used as a starting point:
SELECT *
WHERE {
?question <https://id.parliament.uk/schema/writtenQuestionIndexingAndSearchUin> ?qnum .
?person <https://id.parliament.uk/schema/askingPersonHasQuestion> ?question .
?question <https://id.parliament.uk/schema/questionText> ?text .
?question <https://id.parliament.uk/schema/questionAskedAt> ?date .
FILTER (?date >= "2023-01-01+00:00"^^xsd:dateTime && ?date < "2023-10-01+00:00"^^xsd:dateTime)
}
This query returns the following data:
?question Entity representing a question
?person Entity representing the person who asked the question
?qnum Numerical identifier for this question
?text The text of the question
?date The date the question was asked
Your report must:
1. Document how data was collected, and briefly state any relevant modules or toolkits that were
used. If using SPARQL queries not contained in your code (other than the one provided), you
must include these as separate files within your code submission.
2. Describe your implementation and the models constructed.
3. Summarize your results in appropriate ways, including use of suitable visualizations.
4. Critically evaluate the models. This must involve critical analysis of the adequacy of the modelling
(e.g. what are the assumptions the models rely on, and do they all hold; are there factors or
biases that might invalidate conclusions drawn?), but should also involve some comparison with
external data and/or published social/political science research on relevant subjects.
5. Discuss what conclusions can be drawn from the