Bringing Citizens’ Opinions to Members of Parliament
1. Bringing Citizens’ Opinions to Members
of Parliament
The Newspaper Story
Ruxandra Geana
Steve Taylor
Timo Wandhoefer
CeDEM2012, Donau Universität Krems
2. Contents
• Background & Context
• Problem Statement
• Solution Strategy
• Results
• Conclusions & Further Work
4. Context:
WeGov Project
• Allow policy makers to interact directly with
citizens using Social Network Sites
– Use the tools the citizens already use
• Find and understand
people’s opinions
• Become part of the
discussion
• Open dialog
• Respect privacy
• Encourage trust
7. Scenario
• Policy makers want to test policy statements
with citizens
• Direct contact on SNS can be inefficient
• Often policy statements are issued as press
releases
– Newspaper articles written about them
– These are discussed across social media
– We need to find where the comments are made on
policy statements
• We can then use other tools in the toolbox to
collect postings and provide further analysis
10. Research Questions &
Answers
1. How can we find out where a 1. Google for the headline and URL
news article is being discussed? of the article
2. How can we track the locations for 2. Schedule repeated automatic
1) over time? searches and store ranked results
3. How can we track a news story 3. Group searches into stories
containing many news articles?
4. How can we find news articles 4. Search specific newspaper sites
related to a press release or an for press release
MP’s statement?
5. Which are the important places a 5. Group searches into subject areas
policy maker needs to monitor for & record top tens
a particular subject area?
11. Search Strategy
• Google searches for news article headline & URL
return sites that reference our news article
• Google is by far the most popular search engine
in the Western world
– therefore it is very likely to be used by many people
who may want to comment on a news article
• Google’s ranking is important
– it scores pages by relevance and the number of
citations
– therefore popular pages that reference our news
article will be ranked highly
12. Basic Components
Search Google
for the articles’
Find news URL & headline
articles Configure
Scheduler to
repeat search at
defined intervals
News Article Search URL + Headline Scheduler
User
Data Selection
Internet
Google Custom Search
Database Date: ???
Display & Data Results:
Headline Search
1 www L
Analysis 2
3
www G
www M
Story, Press Release, Minister’s Statement
99
4
5
www J
www A
...
www
Store ranked
Aggregated Results:
1
2
3
Day 1
www A
www J
www Q
Aggregated Results:
1
2
3
Day 2
www J
www A
www Q
Aggregated Results:
1
2
3
Day 3
www K
www J
www Q
Aggregated Results:
1
2
Day n
www AA
www J Analyse
100
URL Search
1
www
www J
search results
3 www AP 2 www K
in database
4 www B 4 www M 4 www P 4 www AE
5 www W 5 www W 5 www X 5 www AW 3 www F
6 www X 6 www K 6 www A ... 6 www AC 4 www D
results
7 www K 7 www X 7 www S 7 www K
8
9
www L
www M
8
9
www L
www B
8
9
www W
www M
8 www L 5 www E
9 www AM ...
10 www 10 www 10 www 10 www
... ... ... ... 99 www
99 www 99 www 99 www 99 www
100 www 100 www 100 www 100 www 100 www
time
13. Illustration of Searches
Over Time
News Article
Date: 5 June 2011 Date: 7 June 2011 Date: 9 June 2011 Date: 11 June 2011 Date: ???
Results: Results: Results: Results: Results:
Headline Search Headline Search Headline Search Headline Search Headline Search
1 www A 1 www B 1 www G 1 www G 1 www L
2 www B 2 www A 2 www H 2 www J 2 www G
3 www C 3 www C 3 www B 3 www B 3 www M
4 www 4 www G 4 www A 4 www A 4 www J
5 www 5 www 5 www 5 www 5 www A
... ... ... ... ...
99 www 99 www 99 www 99 www 99 www
100 www 100 www 100 www 100 www 100 www
URL Search URL Search URL Search URL Search ... URL Search
1 www D 1 www D 1 www F 1 www F 1 www J
2 www E 2 www E 2 www D 2 www J 2 www K
3 www F 3 www F 3 www E 3 www K 3 www F
4 www 4 www 4 www 4 www D 4 www D
5 www 5 www 5 www E 5 www E
... ... ...
99 www 99 www 99 www
100 www 100 www 100 www
time
• Sites referencing the news article change over time
• “Pop Charts” of Rankings
– sites rise & fall, new entrants
14. Analysis: Selection &
Aggregation
• Data can be analysed in different ways
• Select a subset of data
– e.g news story or subject area, each with many
articles
• how do the locations change over time?
• what is the top location over all times?
• Other examples:
– select all news stories in a subject area
– select all news stories in one week
– select news stories from one publisher
15. News Stories,
Articles & OLAP
• One news story
– Multiple news
articles (z axis)
– Multiple searches
over time per article
(x axis)
• We can “collapse” x
and z axes to get
averages for all
values of that
feature
16. Averaging
• Bayesian average chosen to average chart
positions
– Takes account of number of samples each web
site has
– The more instances a site has, the more sure we
are that its position is accurate
• Avoids spurious end-members
22. Conclusions
• A governmental policy maker can discover where
comments are made on policy statements
– Often news articles are written about policy statements
and the articles are discussed over the internet
• We use scheduled and repeated Google searches for
news articles and collect the results in a database
– We can aggregate analyse search results to produce
ranked tables of sites that reference each news article
• Using data mining techniques such as the OLAP
cube, we can group data in many ways
– We can examine aggregate scores taking into account
multiple datasets, averaging out individual differences