See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/318115213
Liberation of public data: Exploring central themes in open government data
and freedom of information research
Article  in  International Journal of Information Management · December 2017
DOI: 10.1016/j.ijinfomgt.2017.05.009
CITATIONS READS
3 104
2 authors:
Eric Afful-Dadzie Anthony Afful-Dadzie
University of Ghana University of Ghana
24 PUBLICATIONS   97 CITATIONS    21 PUBLICATIONS   65 CITATIONS   
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Eric Afful-Dadzie on 02 October 2018.
The user has requested enhancement of the downloaded file.
International Journal of Information Management 37 (2017) 664–672
Contents lists available at ScienceDirect
International Journal of Information Management
journal homepage: www.elsevier.com/locate/ijinfomgt
Research Note
Liberation of public data: Exploring central themes in open government data MARK
and freedom of information research
Eric Afful-Dadzie⁎, Anthony Afful-Dadzie
Department of Operations and Management Information Systems, University of Ghana Business School, Accra, Ghana
A R T I C L E I N F O A B S T R A C T
Keywords: This paper conducts a comparative literature survey of Open Government Data (OGD) and Freedom of
Open government data (OGD) Information (FOI), with a view to tracking the central themes in the two civil society campaigns. With seeming
Freedom of information (FOI) similarities and a growing popularity in research, the major themes framing research on the two movements
Public data have not clearly emerged. Topic modelling, text mining and document analysis methods are used to extract the
Bibliometrics themes as well as key named entities. The topics are subsequently labeled and with expert guidance, their
Topic modelling
Text mining semantic meaning are provided. The results indicate that the major theme in FOI research borders on issues
relating to disclosure, publishing, access and cost of requests. On the other hand, themes in OGD research have
largely centered on technology and related concepts. The approach also helped in determining key similarities
and differences in the two campaigns as reported in research.
1. Introduction (Frey, 2014). Other independent accounts in the literature also show
significant progress in OGD in the US (Krishnamurthy & Awazu, 2016),
Freedom of Information (FOI) and Open Government Data (OGD), UK (Tinati, Carr, Halford, & Pope, 2012), Taiwan (Wang & Lo, 2016),
are two prominent civil society campaigns championing the course of Spain (Carrasco & Sobrepere, 2015), and a host of many other countries
liberating government controlled data to the people. Both FOI and OGD including local government authorities such as cities and federal states.
primarily seek to make government data progressively free and easily In addition, many global experiences have also been shared of how FOI
accessible (Ubaldi, 2013). Largely, the twin but independent campaigns and OGD are impacting governance particularly in the fight against
of FOI and OGD have been driven by (1) a global call on nations to offer corruption, economic empowerment and the quest for greater citizen
a more accountable and transparent governance (2) a growing trend of engagements (Birkinshaw, 2010; Halstuk & Chamberlin, 2006; Jetzek,
sophistication in citizens’ preferences and choices of government ser- Avital, & Bjørn-Andersen, 2012; Shepherd, Stevenson, & Flinn, 2009;
vices (Holler, 2012; Lau, Patel, Fahmy, & Kaufman, 2014; Van Dooren, US. Senate, 2007Zeleti et al., 2016).
Bouckaert, & Halligan, 2015; Weisberg & Nawara, 2010) and (3) an In the wave of the relative progress, there have also been reports of
opportunity to amend past policies, where data collected by govern- misconceptions, myths, definitional challenges and general obstacles
ment agencies tended to be the exclusive reserve of the state (Yiu, besetting real-world practice by the two campaign groups (Camaj,
2012). The two movements have thus become a global mouth piece of 2016; Evans & Campos, 2013; Gigler, Custer, & Rahemtulla, 2011;
advocacy towards a more open and transparent governance. Over the Hubbard, 2008; Janssen, Charalabidis, & Zuiderwijk, 2012; Schartum,
years, the two campaigns have received considerable traction in the 1998; Zuiderwijk & Janssen, 2014). Such wide ranging experiences of
media as well as in academia (Charalabidis, Alexopoulos, & Loukis, “the good, bad and the ugly” of FOI and OGD, have occasioned nu-
2016). For instance, basic statistics regarding the yearly number of merous research publications covering a range of topics. However, as
Freedom of Information Act (FOIA) requests, downloads, appraisal re- the fields of FOI and OGD continue to evolve, what the central themes
ports, number of workshops and conferences held, together with other are as far as research publications are concerned, have not clearly
specific country initiatives, point to a growing interest among stake- emerged. Furthermore, given that the two campaigns not only share
holders (Whitmore, 2012). A similar trend is seen in yearly OGD re- similarities but differences (Ubaldi, 2013), it is imperative to under-
ports, where accounts by the Independent Reporting Mechanism (IRM) stand how key concepts are unconsciously being framed in publications
of Open Government Partnership (the body responsible for the launch relating to the two notions. In view of this, this paper seeks to de-
of OGD in 2011), indicate a steady progress by most member states termine what the major themes have so far been in relation to the two
⁎ Corresponding author.
E-mail addresses: eafful-dadzie@ug.edu.gh (E. Afful-Dadzie), aafful-dadzie@ug.edu.gh (A. Afful-Dadzie).
http://dx.doi.org/10.1016/j.ijinfomgt.2017.05.009
Received 8 August 2016; Received in revised form 20 April 2017; Accepted 19 May 2017
0268-4012/ © 2017 Elsevier Ltd. All rights reserved.
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
campaigns. It is our view that determining the central message shaping helped to establish national and city-based open government data
the two campaigns, would not only shed light on what key topics define portals around the world. Since the two advocacy groups essentially
each movement, but would also help establish whether the similarities lobby governments around the world into adopting the concepts, first
and the seeming differences between the two concepts naturally emerge visible signs of activity are always in the form of endorsements or de-
especially in research. clarations by sovereign governments. As at 2016, 70 countries have
In the years before and after the launch of FOI and OGD, a con- signed the OGD declaration and are at various actionable stages2 On the
siderable number of academic publications have been authored span- other hand, FOI has at 2015, over 95 countries that have adopted some
ning various topics and issues. However, none so far has attempted to mode of FOI legislation.3
comparatively explore the ‘running’ themes in the two concepts. The Besides similarities between FOI and OGD, differences also exist in
few literature reviews available were conducted separately by only terms of the guiding principles and their approach to liberating public
focusing on either of the two notions. For instance, Attard, Orlandi, data (Ubaldi, 2013). For instance, whereas FOI emphasizes on the
Scerri, and Auer (2015) and Novais, de Albuquerque, and da Silva fundamental human rights of people to information (legal arguments),
Craveiro (2013) conducted separate reviews of literature on open OGD campaigns on the economic and social benefits that potentially
government data whiles Halstuk and Chamberlin (2006) conducted a accrue to people when public datasets and documents are made readily
retrospective analysis of the Freedom of Information Act from 1966 to available to all for use, reuse and subsequent distribution (Janssen,
2006. Mendel (2008) conducted a comparative legal survey but how- 2012). It is the view of OGD advocates that, making public data free
ever only centered on FOI. Furthermore, the methodological ap- and easily accessible, not only improves democratic participation but
proaches adopted in review articles conducted on FOI or OGD are dif- also allows individuals to create new products and services based on
ferent from what this paper proposes. This study is therefore uniquely informed and reliable public data (Janssen et al., 2012; Maude, 2012).
positioned to contribute to theory and fill research gaps in the following The primary goals of the two campaigns, according to Janssen (2012),
ways. First, key topics and associative terms ‘running’ in scientific also influence the kinds of datasets mostly sought after in each group.
discourses on the two civil movements are identified and compared. Datasets such as transport data, geographic data, corporate data and
Secondly, the paper also identifies location-based named entities of general business information which have the potential to spur innova-
interest to frame how FOI and OGD are being implemented around the tion and economic growth, are mostly of interest to OGD advocates. On
world. The results provide a means to understand the central themes the other hand, given the professional background of most FOI propo-
shaping each campaign and some potential future research directions nents (lawyers, media practitioners etc.), the most preferred datasets
on OGD and FOI. tend to be government budgets and expenditure, revenue data, legal
The rest of the paper is organized as follows. First, a brief overview information as well as documents covering reports and meeting minutes
of FOI and OGD covering history and key concepts are explained. of key government agencies. Another dimension of OGD advocacy
Further, differences and similarities between the two notions as cap- different from FOI, is the emphasis on deploying latest information
tured in the literature, are presented. This is followed by the metho- technology tools that help in the release of public datasets. The tech-
dology which explains the approach to data collection and presents a nical dimension of OGD sets it apart from FOI particularly on how clear
brief introduction to text analysis concepts and their relevance to the specifications are recommended on how to model, create, publish, store
study. Research questions guiding the study are subsequently pre- and release government data in various formats for easy accessibility.
sented. The results of the study, discussion and conclusion are further For instance, proponents of OGD tend to emphasize that datasets are
presented. stored and accessed in non-proprietary machine readable formats, such
as CSV, XML, TSV, RDF (Berners-Lee, 2006). Furthermore, OGD re-
2. FOI Vs OGD commends that adequate metadata and linked data technology are
provided for public datasets. Thus, unlike RTI, the concept of open
The idea of open government data (OGD) is generally viewed as an government data (OGD) transcends the emphasis on fundamental
o shoot of Freedom of Information (FOI), also sometimes known as human rights of people to access information. Rather, focus is also onff
Right to Information (Ubaldi, 2013). However, the two movements providing data management and exchange tools that make the idea of
come under the broader concept of open government, which seeks accessibility, re-use and distribution, a reality. Other recognized dif-
transparency and greater rights of information access for citizens ferences between FOI and OGD are also seen in the way they approach
(Tauberer, 2012). It must be noted that while civil resistance move- copyright and licensing issues. FOI implementation tends to differ from
ments are not completely new, the quest for openness and access to country to country especially since some countries are very restrictive
government controlled information (spearheaded by FOI and OGD), is a on what can be published and reused. OGD on the other hand, grants an4
fairly modern concept. Together, the two campaigns are contributing to inherent license to users to freely use, reuse and distribute public data.
opening wide the frontiers of democracy, ghting corruption and em- Further, because of the often legal approach adopted by FOI, its re-fi
powering citizens. lationship with other stakeholders, especially public sector employees
Freedom of Information was o cially given a stamp of approval in a and politicians are sometimes antagonistic rather than collaborative asffi
United Nations General Assembly Resolution in 1946. It was further often seen in OGD approaches.
5 In the following section, the metho-
strengthened in 1948 through Article 19 of the Universal Declaration of dology adopted in the study to investigate the central themes ‘running’
Human Rights (Donnelly, 2013). The UN initiative, set the tone for the in academic publications in the two movements is explained.
adoption of FOI Act (FOIA) in most countries. Open government data on
the other hand, was first conceived in 2007 when 30 OGD campaigners 3. Methodology
met in California to deliberate on principles to guide the new movement
(Chignard, 2013). This meeting and other close efforts such as the The main part of the research design used topic modelling. Text
Public Sector Information (PSI) Directive of 2003, U.S. President mining and document analysis methods were used mainly to clean and
Obama’s open data initiative of 2009, helped set up the open govern-
ment data partnership (Attard et al., 2015; Tauberer, 2012). More re- 2 http://www.opengovpartnership.org/countries.
cently, the G8 Open Data Charter.1 in 2013 have all contributed to 3 http://www.right2info.org/access-to-information-laws/access-to-information-laws#_
strengthening the ideals of OGD. These current and past initiatives have ftnref7.
4 http://webfoundation.org/2015/08/freedom-of-information-and-open-government-
data-communities-could-benefit-from-closer-collaboration/.
1 http://opendatacharter.net/history/. 5 https://www.article19. org/data/files/pdfs/standards/righttoknow.pdf.
665
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Table 1
Inclusion and Exclusion Search Criteria.
Search Criterion/Task Inclusion Exclusion
Comparative Survey FOI; OGD Any unrelated
Document Type Journal articles, Conference proceedings and Book chapters. Editorial, Doctoral dissertations, master’s theses, textbooks, Letters, Erratum etc.
Language English Non-English
Time Period FOI: – 2015; OGD: – 2015 After 2015
Bibliographic database Web of Science and Scopus All other database
Data cleaning Main text spanning introduction to conclusion sections of each Title page, keywords, references, funding sources and acknowledgements
article sections.
transform the textual data. The methodology is conveniently segmented
into three phases of text pre-processing, processing and information ex-
traction as shown in Fig. 2. The three stages were however preceded by
a data collection phase which primarily employed document analysis
(Owen, 2014) techniques to gather the data. Particularly, the inclusion
and exclusion search criteria in document analysis was used to refine
the data search as shown in Table 1. The following section presents a
brief introduction to the text analysis approaches used in the study and
further explains what the data collection stage entailed.
3.1. Text analysis
The era of massive generation of digital textual data is impacting
positively on text analysis research and practice. Robust tools and
methods (Aggarwal & Zhai, 2012) that help to glean meaningful in-
formation from large amounts of documents are consistently being in- Fig. 1. The Latent Dirichlet Allocation (LDA) Model.
troduced to keep pace with a growing research field. Text analysis tools (Source: Blei et al., 2003)
and methods are popularly used in information and document retrieval,
text summarization, document dis(similarity) identification, language
identification and document authorship attribution (Witten, 2005). that wv = 1 and wu = 0 for u≠ v. The following definitions are used to
However, irrespective of the text analysis method involved, the end further explain the concept as demonstrated graphically in Fig. 1. A
goal is always about extracting high-quality information from textual document is defined as a sequence of N words expressed as
data. An effective text analysis project, according to DiMaggio, Nag, W = (w1, w2,…,wN), where wN is the nth word in the sequence.
and Blei (2013), must (1) be reproducible (2) automated (3) inductive A corpus is a collection of M documents denoted by
to patterns and (4) provide “relationality” in different contexts. These D = (w1, w2,…,wN)
requirements demonstrate an improvement over non-computational α is the parameter of the topic Dirichlet prior per document topic
methods. In this study, the term text analysis is expediently used to distributions
encompass digital text computational methods such as text mining and β is the parameter of the word Dirichlet prior per document topic
topic modelling, even though the terms text analysis and text mining distributions
are also sometimes used synonymously. In the following section, a brief θ is the topic distribution for document i
introduction to topic modelling as used in the study is presented. φ is the word distribution for topic k,
D = {wi} corpus, each wi denotes a word
3.1.1. Topic modelling Z = {zi} latent topic assigned to words in W
In the last decade, the concept of topic modelling has become one of Over the years, topic modelling has been used extensively in a
the most trending research areas in text analysis (Wang & Blei, 2011; number of areas such as fraud detection (Xing & Girolami, 2007), spam
Wei & Croft, 2006). Topic modelling is basically used to summarize filtering (Bíró, Szabó, & Benczúr, 2008), clustering scientific documents
large corpora of text by revealing hidden themes in textual data (Blei, (Yau et al., 2014), determining business topics in a source code
Ng, & Jordan, 2003; Chang, Gerrish, Wang, Boyd-Graber, & Blei, 2009; (Maskeri, Sarkar, & Heafield, 2008), detecting phishing websites
DiMaggio et al., 2013). There are three topic models namely Latent (Ramanathan &Wechsler, 2012), framing ideas in history (Hall et al.,
Dirichlet Allocation (LDA), Probabilistic Latent Semantic Indexing and 2008), social network analysis (Ríos, Aguilera, Bustos,
Correlated Topic Models. The most utilized model is the LDA algorithm, Omitola, & Shadbolt, 2013), understanding how the media frames cer-
which uses Bayesian statistical model to treat documents as a random tain topical issues (Afful-Dadzie, Nabareseh, Oplatková, & Klímek,
bag of words over latent topics and where each topic is characterized by 2016) among many others. Topic modelling applications have also been
a distribution over words (Blei et al., 2003; Blei, 2012). Further, the extended to bibliometric review, as evident in the following articles
technique works by identifying clusters of words that frequently co- (Griffiths & Steyvers, 2004; Jiang, Qiang, & Lin, 2016; Mann,
occur in a corpus of text without regard to word order. In view of this, it Mimno, &McCallum, 2006). The use of text analysis in bibliometric
is recommended that a relatively large size of textual data is used so as analysis is relatively new and the approach is a departure from manu-
to help increase the likelihood of themes of text being found together. ally analyzing documents one at a time. In this study, LDA based topic
The LDA algorithm may be described formally as shown in Fig. 1 culled modelling, general text mining techniques and document analysis are
from Blei et al. (2003) and Yau, Porter, Newman, and Suominen (2014). used to extract information from research articles with a view to es-
In a vocabulary indexed as {1,…,V }, words are represented as unit- tablishing central themes that conceptually frame research discourses
basis vectors where a unit document is equal to one and the rest of the on open government data (OGD) and Freedom of Information (FOI).
components equal to zero. Representing the components with super- The following section explains the data collection phase as simplified in
scripts, a vth word in the vocabulary is denoted by a V-vector w such the inclusion and exclusion search criteria in Table 1.
666
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Fig. 2. Conceptual framework for the com-
parative survey.
(Source: authors)
4. Data collection research questions. In all, the questions were designed to aid in
‘framing’ trending concepts in FOI and OGD as captured in research
Research publications were selected guided by the following criteria publications. The following questions guided the research:
(i) an automated search availability (ii) quality and prominence of RQ1. What are the central themes in FOI and OGD research pub-
publications and (iii) reputation of the bibliographic database. In line lications?
with the above considerations, we settled on the Web of Science and RQ2. How do similarities and differences identified in the topics
Scopus; widely recognized as the two most prominent bibliographic compare with what have specifically been written in the literature?
databases (Aghaei Chadegani et al., 2013; Wang &Waltman, 2016). RQ3. How do the topic labelling (classification) frame the discourses
Journal articles, conference proceedings and book chapters were the of FOI and OGD research?
only kinds of research documents deemed appropriate for the study. RQ4. What do named entities especially occurrences of countries,
The document search was aided by relevant key search phrases and regions and cities in the topics say about the campaign?
terms, which ensured that only articles suitable to the study were se-
lected. For instance, search phrases such as ‘freedom of information’,
5.1. Phase 1: pre-processing
‘freedom of information act’, ‘right to information’, ‘right to informa-
tion act’, ‘access to information (ATI)’ were used to search articles on
The pre-processing phase involved a number of text cleaning pro-
FOI. Similarly, ‘open government data’, ‘open data’, ‘government data
cedures to transform the documents into requisite formats for proces-
portal’, ‘government public data’ and ‘public data portal’ guided the
sing. The first procedure utilized a series of R programming codes to
search of articles on OGD.
convert the pdf documents into text format. Subsequent codes helped to
In addition, a pre-condition was that, key terms or phrases must
strip the articles of title pages, abstracts, keywords, references, funding
have appeared in the article title but without strict recourse to word
sources, notes, acknowledgements and appendixes, leaving only the
sequencing. This meant that terms such as ‘government open data’ or
‘main’ content beginning from “introduction to conclusion”. This pro-
‘public government data’ were considered as guides for selection. The
cedure was necessary to eliminate sections in the articles that could
search strategy was partly adopted from Attard et al. (2015). After in-
unduly influence the outcome of the text analysis. Following this, a
itial downloads of the articles, a document analysis was carried out to
number of steps were taken to transform the corpus ready for proces-
further scrutinize the articles. This approach ensured that only articles
sing. These included converting the corpus into lower case characters,
related to the concepts under study were included. For instance, there
striping whitespaces, removing punctuations, sparse terms, numbers as
were article titles that had the phrase ‘right information’ but which
well as stop-words. The study utilized relevant packages in the R
actually had nothing to do with right to information. A thorough
Statistical Computing Software for all the text analysis including the
document analysis helped to remove all such unrelated articles. In all,
pre-processing and the text processing phases. The main libraries or
780 articles were collected on FOI and were subsequently trimmed
packages utilized in the study were the tm package (for text mining),
down to 430. Similarly, the OGD article search resulted in 392 articles
mallet package (for LDA topic modelling), and SnowballC (for stem-
of which 281 were deemed appropriate for the study.
ming).
5. Research questions 5.2. Phase 2: text processing and results
To get the most out of the comparative literature survey, a set of This stage involved the deployment of the mallet LDA library to-
pre-defined questions were used to guide the study. This approach was gether with other relevant libraries in R to first train, and subsequently
particularly useful as it helped to map the results generated to the generate the topics. The study experimented with a different number of
667
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Table 2
The Top 15 frequent words in each topic on FOI research.
Topic Frequent Terms (Stemmed)
Topic 1 privaci- mug public shot court interest disclosur. u.s. exempt foia feder- circuit crimin- law report
Topic 2 social societi- state polit- economy- knowledg- role process develop technolog- market mean concept model peopl-
Topic 3 freedom inform law media law press foil polit- news diffus- legisl- nation studi- country- access
Topic 4 health data clinic trial right public registr- supra note trial human privaci- drug report research
Topic 5 rti govern india public rtia state corrupt citizen peopl- local indian delhi act societi- time
Topic 6 request foi inform govern institut- feder- public access law canada law request provinci- commission atia
Topic 7 media foia news public journalist request fee waiver blogger interest request blog document interest context
Topic 8 librari research africa servic- world univers- develop local- servic- interest nation internet report communic- commiss-
Topic 9 ati/foi belief request govern belief agenc- account evid- autonomi- text work request truth agent inspect
Topic 10 right human inform internet world intern organis- communic- onlin- group activist countri- wikileak communic- work
Topic 11 foi countri- act south implement countri- africa public request law societi- studi- level develop civil
Topic 12 right court articl- adolesc- human european ecthr convent express posit public case media societi- state
Topic 13 from made time case import number part nation case issu- system includ- year found make
Topic 14 act inform public interest author commission foi request subject decis- code request exempt disclosur- held
Topic 15 foi govern cabinet disclosur- request freedom legisl- polit- account civil reform impact london flow central
Topic 16 act govern public agenc- foia record feder- agenc- congress execut- secur- law hous- u. s offic-
Topic 17 inform access freedom legal protect data make law person privat- provid- general polici- citizen privaci-
Topic 18 court agenc- foia document request disclosur- from exempt record district depart subject cir appeal u. s. c
Topic 19 corrupt foia state public offici- court strong from state year convict law local govern law
Topic 20 govern public european document council parliament bill institut- author offici- legisl- provis- open institut- administr-
Topic 21 data record manag- research request foi public request author legisl- local council nhs studi- vexati-
Topic 22 public law commiss- inform decis- disclosur- court regul- articl- applic- bodi- administr- decis- ministry- foi
Topic 23 state oil nepa pipelin- environment keyston- sand public depart climat- emiss- chang- propos- impact ghg
Table 3
The Top 15 frequent words in each topic on OGD research.
Topic Frequent Terms (Stemmed)
Topic 1 contract qualiti- busi- portal result gdp averag- number countri- dataset tabl- reliabl- publish offici- notic-
Topic 2 govern servic- citizen app model data citizen provis- develop platform busi- crm mobil- citi- cultur-
Topic 3 govern progress web figur- social brazil websit- feder- analysi- government websit- societi- public peopl- system
Topic 4 govern transpar- matur- level open agenc- social model collabor- particip- benchmark engag- media feder- benchmark
Topic 5 dataset link ogd dataset architectur- sourc- social approach initi- portal integr- publish order metadata model
Topic 6 logd web portal dataset entiti- data. gov sourc- rdf integr- contract u. s databas- entiti- agenc- sourc-
Topic 7 visual catalog ogd catalog data portal user display system visual tool analysi- qualiti- web record
Topic 8 ogd innov- adopt benefit barrier factor organ perceiv- busi- influenc- social janssen busi- model user
Topic 9 govern open agenc- govern initi- transpar- polici- dataset portal develop countri- particip- share local privat-
Topic 10 india organis- nation polici- technolog- rti peopl- studi- act state e − govern indian societi- govern centr-
Topic 11 inform public servic- research citizen access provid- case govern make knowledg- onlin- work sweden relat-
Topic 12 busi- model municip- citi- urban user social set model group citi- dimens- dimens- strategi- privat-
Topic 13 public inform sector psi re − use direct european access polici- licens- bodi- govern licens- principl- australian
Topic 14 project releas- aid develop polici- local competit- research citi- transpar- project approach effect plan agenda
Topic 15 mechan- ogd generat- social innov- economy- energi- access open effici- resourc- particip- technic sector opow
Topic 16 ogd citi- survey implement question research polit- result depart econom- questionnair- benefit administr- respond onlin-
Topic 17 link rdf web semant- sparql dataset time ontolog- queri- result metadata forest fire servic- work
Topic 18 ogd research capabl- level dimens- user methodolog- platform evalu- measur- variabl- layer generat- model domain
Topic 19 data open base process analysi- import level user set exist framework result model nation qualiti-
topics to arrive at the ideal number appropriate for each dataset 6. Topic interpretation
(corpus). This was done by generating posterior likelihoods on a
number of models assigning different numbers of topics each time. The FOI
result indicated that a maximum of 23 and 19 topics were respectively The topic labelling or classification was done by interpreting what a
appropriate for the FOI and the OGD corpora. The results of the topic body of topics appears to convey. Guided by expert knowledge and the
modelling are as presented in Tables 2 and 3 respectively for FOI and literature, it was realized that a number of the topics seemed to fall
OGD. under some relevant issues in the two campaigns. In row 1 in Table 4
for instance, the topic label apparently frames issues relating to some
FOI guiding principles and key operational terms. This is because, most
5.3. Phase 3: information extraction authoritative texts on FOI particularly those that focus on Article 19
(Birkinshaw, 2010; Foerstel, 1999; Mendel, 2008), tend to recognize a
The topics were interpreted based on expert knowledge and support set of guiding principles that are fundamental to the movement.
of the literature. After generating the topics, they were subsequently These FOI guiding principles are “maximum disclosure, obligation to
labelled to help frame the central themes in the two research domains publish, promotion of open government, limited scope of exceptions (ex-
as shown in Tables 4 and 5 respectively. The following sub-sections emptions), processes to facilitate access, costs, open meetings, disclosure
present relevant information that were extracted to give meaning to the takes precedence and protection for whistleblowers”. The topic label dis-
topics under each of FOI and OGD. Further, the results from FOI and closure/publishing/access/costs which has 40.0% of the topics, captures
OGD are compared for similarities, differences and general trends in the most of the above FOI principles. In addition, row 2 in Table 4 frames
topics. topic scenarios that have semblance to law, legislation and exemptions to
668
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Table 4
FOI topic labelling and concept framing.
Concept framing (Topic labelling) Topic Topic proportion (%)
Disclosure/Publishing/Access/Costs Topic 1, Topic 5, Topic 6, Topic 7, Topic 9, Topic 10, Topic 15, Topic 16, Topic 18, Topic 21 40.0
Law/Legislation/Exemptions Topic 3, Topic 12, Topic 14, Topic 17, Topic 18, Topic 19, Topic 20, Topic 21, Topic 22 35.0
Service provision Topic 2, Topic 8, Topic 11, Topic 21, 15.0
Health/Environmental Topic 4, Topic 21, Topic 23, 10.0
Table 5 In the health/environment category, two topics, 4 and 23, capture
OGD topic labelling and concept framing. the discourses around health and environmental information dis-
closures. Topic 4 with words like “health”, “data”, “clinic”, “drug”,
Concept framing (Topic Topic Topic proportion
labelling) (%) “privaci-”, “report” and “research” is similar to Topic 21 and appears to
focus on health information disclosures, access and attendant privacy
Transparency/Collaboration/ Topic 4, Topic 8, Topic 9, 12.5 issues. Topic 23 is loaded with lots of environmental issues and con-
Participation Topic 14, cerns relating to freedom of information. The terms “pipelin-”, “keyston-
Technology Topic 2, Topic 3, Topic 5, 50.0
Topic 6, Topic 7, Topic 17, ”, “oil” and “nepa” readily bring to mind the oil pipeline system in
Topic 18, Topic 19 Canada and the United States, the National Environmental Policy Act
Economic/Social/Innovation Topic 1, Topic 8, Topic 10, 18.75 (NEPA) and related concerns. The other terms in the category like
Topic 14, Topic 15, Topic 16 “environment”, “climat-”, “emiss-”, “chang-”, “impact” and “ghg” appar-
Citizen engagement Topic 11, Topic 12, Topic 18.75 ently address issues about how the keystone pipeline project would
13, Topic 14, Topic 16
impact climate change through emissions and related concerns about
greenhouse gas (GHG) emissions. The study further found that most of
FOI requests and disclosures. The law, as already explained in the in- the key terms in the various topics happen to have a strong correlation
troduction, is a major driving force behind the FOI campaign. Finally, among themselves. Terms such as ‘access’, ‘media’, ‘exempt’, ‘legisl-’,
rows 3 and 4 label topics that cover a number of other key issues such as ‘privaci-’, ‘request’ and ‘law’ had at least a correlation of 0.70 among
FOI service provision, and health and environmental issues respec- themselves. This gives credence to their regular occurrences in the to-
tively. pics. Topic 13 seemed to be an outlier as no relevant interpretation
It can be recognized that Topics 1, 5, 6, 7, 9, 10, 15, 16, 18 and 21 could be drawn from it. In determining the proportion of topics under
embody themes that relate to most of the key principles in FOI. For each label, Topic 13 is not counted as well as topics that appeared in
instance, Topic 1 addresses the issue of access to mug shots information more than one category, such as Topics 18 and 21.
which so often raises conflicting issues about privacy on one hand, and Overall, we recognize from the proportion of topics under each
access, disclosures and exemptions on the other hand. Topic 1 specifi- classification that the central themes in FOI research have largely been
cally points to the many instances of legal tussles fought in US. courts on issues relating to the core principles of FOI; be it issues surrounding
regarding whether or not there should be exemptions to requests for disclosures, publishing, access, costs, exemptions among others.
mug shot booking information. Terms such as “privaci-”, “mug”, “in- OGD
terest”, “disclosure”, “exempt”, “crimin-” and “law” capture the senti- Since the launch of open government data in 2009, the campaign to
ments surrounding the issue particularly the controversy about some liberate public data has largely been shaped by the terms Transparency,
companies profiting from the sale of mugshot photos (Rostron, 2013) – Openness, Participation and Collaboration (Krishnamurthy & Awazu,
an attack on individual privacy. While Topic 6 focuses on general FOI 2016; Lathrop & Ruma, 2010; McDermott, 2010; Veljković,
requests and the law at federal and provincial levels in Canada, topic 7 Bogdanović-Dinić, & Stoimenov, 2014). Surprisingly however, many of
with terms such as “media”, “journalist”, “fee”, “waiver” and “blog”, the topics apparently do not capture these widely accepted pillars of
appears to be addressing the issue of costs to requesters of freedom of OGD. As shown in Table 5, only two topics representing 12.5% of the
information, as evidenced in many countries including the US7. Topics total topics implicitly talk about such. Other labels identified in the
9, 10, 15, 16 and 18 also carry similar themes about FOI requests, topics were Technology; Economic/Social/Innovation and Citizen engage-
disclosures, access, politics and the reach of the law regarding FOI ment. As many as 50.0% of the topics centre on technology related
implementations. Topic 21 with words such as “data”, “research”, “re- matters. This seems to affirm the position by Janssen (2012) that OGD
quest”, “nhs”, “public” and “vexati-” seems to be addressing general unlike FOI/RTI is technology driven rather than a rights seeking cam-
frustrations (vexations) regarding access to National Health Service paign. For instance, Topic 2 contains the terms “app”, “crm”, “mobil”,
(NHS) data even for the purposes of research. Though no specific “citizen” and “platform” suggesting the deployment of technologies such
country is mentioned in Topic 21, there appears to be a general ap- as mobile apps, customer relationship management (CRM) solutions
prehension among stakeholders regarding the use and re-use of sensi- and other platforms to engage citizens in a democracy. Topic 5 is rich
tive clinical data under the FOI Geissbuhler et al. (2013), posits that with OGD technology related terms such as, “link”, “ogd”, “dataset”,
while concerns about access to clinical data is genuine and must be “architectur-”, “portal” and “metadata” which appear to rehash the
regulated by governments, sharing health data advances public health technology expectations of a functional open government data plat-
research and improves patient care. form.
In the category of law/legislation/exemptions, Topics 3, 12, 14, 17, Topic 6 contains other unique technological terms like “logd (linked
18, 19, 20, 21 and 22 frame a general theme around the law, legisla- open government data)”, “web”, “rdf (Resource Description
tion, the courts, and privacy protection. Topic 12 in particular mentions Framework)”, “databas-” and “entiti-” which readily point to calls by
“ecthr” which is the European Court of Human Rights and therefore OGD proponents for structured non-proprietary machine-readable
appears to be case settlements regarding FOI. Topics 19 and 22 under technologies that support linked data and also aid in data access use
the same category have terms like “corrupt”, “state”, “public”, “offici-”, and redistribution. Topic 7 under the same category also mentions
“convict”, “commisi-”, “court” and “legisl-” appear to be raising issues of several technology related terms. It can further be seen that Topic 17 in
corruption and the need for the law to prosecute state officials im- particular focuses on semantic ontology technologies as evidenced in
plicated in corrupt deals. terms such as “rdf”, “sparql”, “semant-”, “ontolog-”, “queri-” and
669
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Table 6
Key named entities identified in the various topics.
Topics Name entity: country/region/ Other Entities
city
FOI/RTI Topic 1, Topic 5, Topic 6, Topic 8, Topic 11, Topic 15, Topic 16, U.S, India, Canada, Africa, Politicians, citizen(s), media, journalist, wikileak, court, parliament,
Topic 20; Topic 2, Topic 7, Topic 10, Topic 18, Topic 23 London, Europe nhs (National Health Service), key stone pipeline
OGD Topic 3, Topic 4, Topic 6, Topic 10, Topic 12, Topic 13; Topic 2, Brazil, U.S, India, Sweden, Citizen(s), media, user, janssen
Topic 8, Europe, Australia
“metadata”. Semantic ontologies add value to an otherwise unconnected continue to charge fees to administer FOI requests (Goodspeed, 2011).
data by offering endless possibilities with linked data such as utilizing it Again, in spite of the fact that FOI charges fees, the process of obtaining
across platforms and organizations. Linked data also help in generating an information is also comparatively laborious than with OGD. For
data analysis, reports, publications and maps (Geiger & von Lucke, instance a requester may have to first identify a relevant agency that
2012; Shadbolt et al., 2012). Topics 18 and 19 also appear to mention has the information, compose a formal letter, pay a request fee and
technological terms all related to open government data. follow up on the application.6 For the most part however, OGD re-
Geiger and von Lucke (2012) argue that, beyond the often nar- questers only have to access the needed data directly from national or
rowest definition of OGD which revolves around the terms Transpar- city-based dedicated web portals in various data formats at no cost.
ency, Openness, Participation and Collaboration, other terms that matter The comparison also brings to the fore similarities especially with
are social innovations and economic development. In this regard, Topics 1 names of recognized stakeholders under the two campaigns. The FOI
and 15 implicitly capture such themes in the discourses of open gov- topics make mention of names such as journalist, politicians, public of-
ernment data. With terms such as “gdp”, “reliabl-”, “publish”, “busi-” and ficials, citizens and by inference lawyers (from the many occurrences of
“countri-”, Topic 1 mildly seems to address the socio economic dimen- the terms, “laws” and “courts”). The topics under OGD also mentions
sion of OGD and its inherent potential to trigger business innovations entity names like government, media and citizens which come close to
among citizens (Janssen et al., 2012; Maude, 2012). The themes in those identified in the FOI topics. This may seem to suggest that the two
Topic 15 also appears to reinforce the social, economic and innovation campaigns have a similar core group of stakeholders.
dimensions of OGD. Another concept often talked about in relation to
OGD, is citizen engagement. Topics 11, 12, 13 and 16 contain words 6.2. Named entities: country/regional/city-based initiatives
that seem to describe some form of government-citizen engagement.
The study further investigated whether some of the key terms as Named Entity Recognition (NER) is an information extraction task
identified in the topics correlate strongly among themselves. Terms that that seeks to identify and subsequently classify noun phrases (entities)
are strongly associated with at least a 0.70 correlation measure were found in textual data (Downey, Broadhead, & Etzioni, 2007). Typically,
‘access’, ‘media’, ‘exempt’, ‘innov-’, ‘transpar-’, ‘link’, ‘ogd’, ‘technolog-’, the noun phrases may refer to persons, locations, organizations, time,
‘dataset’ and ‘format’. Overall, the proportion of topics indicate that the money or any other entity of interest. In this study, the topic modelling
major central theme in OGD research have so far centered on tech- identified some key entities particularly location-based entities that
nology and related issues. So much has been written on data publishing may be worthy of mention. These entities under FOI and OGD are
standards and technologies especially as relating to data formats, open classified as named entity: country/region/city as shown in Table 6. It
linked data, the architecture and general data management standards as must be noted that while the named entities alone would not reveal a
used in open government data portals. The following section compares complete information, they often represent an activity, progress, im-
the key themes identified in each campaign for similarities and differ- pediment or an initiative relating to either FOI or OGD.
ences. Particularly, we find out whether there are any similarities or In FOI research, some location-based (countries, regions and cities)
differences not already mentioned in popular literature. entities can be recognized in the topics as observed in Table 2 and
subsequently classified in Table 6. Some of these entities are the U.S,
6.1. Topic comparison India, and Canada (in Topics 1, 5 and 6 respectively); Africa and Europe
(respectively in Topics 8 and 12); Delhi and London (in Topics 5 and 15).
The topics extracted and their apparent meanings bring out several In the case of the US as a named entity, we recognize that while nu-
familiar themes in FOI and OGD research. They also further reveal some merous publications address a range of topics, most have centered on
similarities and differences between the two concepts as discussed the core issue of ‘how much’ information can be disclosed. This issue is
earlier. For instance, the topics implicitly affirm that the FOI campaign often contested in various levels of the US judicial system and in con-
has largely thrived on law whiles much of OGD’s focus has been about gressional hearings (Relly & Schwalbe, 2016). On India, much of the
open data technologies. The identified major theme in OGD research FOI articles have focused on the law, access, and the impact of FOI on
confirms the fear by Geiger and von Lucke (2012) that, care must be
fights against corruption in public office (Calland & Bentley, 2013;
taken to not narrow or reduce the focus of OGD to data access tech- Roberts, 2010). Other entities refer to key FOI stakeholders such as the
nologies. Some named entities identified in the sets of topics under FOI media, journalists, politicians, citizen, parliament, courts, library, etc. The
and OGD, also appear to reinforce the belief that OGD has for the most National Environmental Policy Act (NEPA), WikiLeaks are also some of
part concentrated on the use of technology whiles FOI focuses on the the named entities mentioned in FOI topics as shown in Table 2. Si-
law. For instance, terms (names) that refer to an individual involved in milarly in OGD research, various entities mostly location-based can be
an FOI activity, tended to use words such as citizen, public, person and identified in the topics generated. These are Brazil, U.S, Sweden, Aus-
author as seen in Table 4. However in the OGD topics, the term “user” tralia, India and Europe. The mention of Sweden in Topic 11 in Table 3 is
which is predominantly used to describe a kind of computer or in- not particularly surprising since most texts that trace the history of open
formation systems operator such as “end-user” frequently appeared. government often make reference to Sweden as one of the first coun-
The topics and their subsequent labelling further reveal that, unlike tries to pass a legislation over 200 years ago, to make public
OGD that stresses on a free at no cost public data, data under FOI
cannot be said to be entirely free since requesters in most countries
including the U.S, U.K, Ireland, Australia, Canada and Scotland 6 https://www.icij.org/resources/2012/04/freedom-of-information.
670
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
information accessible to citizens (Janssen, 2012; Mendel, 2008). One data free at no cost, FOI requesters are still charged requests fees even
other unique named entity appears in Topic 8 by way of “Janssen”. This in many advanced countries. On the other hand, OGD data sets appear
may be so because Katleen Janssen and Marijn Janssen have con- to be completely free of costs to citizens who have an added luxury of
tributed so much to research in open government with lots of citations accessing the materials in many formats. It is not clear whether the
between them. Two of their most cited articles are “Benefits, adoption kinds of data requested under FOI warrants the charges. The study also
barriers and myths of open data and open government” by Marijn and shows that, comparatively, accessing an OGD data is easier than the
“The influence of the PSI directive on open government data: an often strenuous processes that requesters go through to obtain FOI data.
overview of recent developments” by Katleen. In terms of the topic labelling, research question 3 is answered since the
use of expert knowledge and the literature helped to frame the dis-
7. Discussion and conclusion courses surrounding the two subjects and subsequently extracted va-
luable information that give a general trend of the themes in each
Several decades have passed since the freedom of information act campaign.
(FOIA) was conceived as a means to providing access to public data and Research question 4 sought to understand how named entities
with a view to entrenching the values of democracy. After many years identified in the topics comment about each campaign. The key named
of global successes and challenges in implementation, a similar move- entities give an indication for further research. For instance, a thorough
ment in the form of open government data (OGD) was launched to help review of publications revealed that, whiles numerous country-level
support the idea of greater openness and accountability in governance. FOI activities have been reported in the literature, only a few have
Though run independently, the two campaigns continue to draw the focused on country-to-country comparisons. It would be interesting to
world’s attention to the importance of establishing a free accessible see future research focus on comparing countries over some FOI per-
public data regime to augment the values of democracy. While the two formance measures such as the level of involvement by civil societies
concepts continue to receive considerable attention, the major themes and the media, the impact of FOI on corruption fights among others.
that run in scientific written discourses have not clearly emerged. In Though this approach should not be intended as a score card on
this paper, topic modelling, text mining and document analysis countries, it would give a general trend of FOI performances world-
methods were harnessed to determine the major topics running through wide. Similarly, we also realized that whereas much has been written
FOI and OGD research publications and to establish whether these about individual country performances, initiatives and general hap-
central themes help frame the ideologies in the two concepts. penings on OGD, research seems to be silent on city or state based ac-
The text analysis approach used in this comparative bibliometric tivities. At present, only a few OGD city-based scientific publications
analysis is a departure from traditional approaches where manual have been written such as on Chicago, Vienna, Rotterdam, Bologna and
document analysis methods are used in extracting the metrics. In ad- Trentino. This appears to give OGD a narrow scope since open gov-
dition, traditional literature review approaches habitually tends to ernment data is not only meant to be practiced at the level of central
focus on information such as yearly number of publications, citation governments. For instance, while over 70 international countries are
index, leading authors and affiliations, top journals and domain areas. involved in OGD programmes, there are also 164 international cities
Text analysis methods are now providing a computational alternative to and regions practicing OGD with independently run data web portals.2,7
bibliometric analysis through document authorship attribution, lan- In view of this, future research on OGD should focus on various city-
guage identification, document retrieval and clustering. based activities to help broaden the scope and the understanding of
The topic modelling not only helped in establishing the central what OGD truly entails.
themes in FOI and OGD research but also helped to clearly define si-
milarities and differences in the two campaigns. The topic classification References
was carried out to determine the proportion of topics that fall under
each of the labels. Since the topic labels mostly reflected key concepts Afful-Dadzie, E., Nabareseh, S., Oplatková, Z. K., & Klímek, P. (2016). Framing media
in each campaign, a topic label with a significant proportion of the coverage of the 2014 sony pictures entertainment hack: A topic modelling approach.
Proceedings of the 11th international conference on cyber warfare and security:
overall topics, give an indication of the central theme in that subject. In ICCWS2016. Academic Conferences and publishing limited (p. 1).
this respect, the results indicated that the central theme in FOI research Aggarwal, C. C., & Zhai, C. (2012). Mining text data. Springer Science & Business Media.
have largely centered around issues of disclosure, publishing, access and Aghaei Chadegani, A., Salehi, H., Yunus, M. M., Farhadi, H., Fooladi, M., Farhadi, M.,
cost of requests. The next major theme identi ed in FOI research hovers et al. (2013). A comparison between two main academic literature collections: Webfi of science and Scopus databases. Asian Social Science, 9(5), 18–26.
around issues of the law, legislation and exemptions relating to the FOI Attard, J., Orlandi, F., Scerri, S., & Auer, S. (2015). A systematic review of open gov-
act. These two themes closely reflect most of the guiding principles in ernment data initiatives. Government Information Quarterly, 32(4), 399–418. http://
FOI as particularly enshrined in Article 19 of the Universal Declaration dx.doi.org/10.1016/j.giq.2015.07.006.
Bíró, I., Szabó, J., & Benczúr, A. A. (2008). Latent dirichlet allocation in web spam fil-
of Human Rights. On the other hand, the major theme in OGD research tering. Proceedings of the 4th ACM international workshop on adversarial information
as shown in the topics seems to center on technology and related sub- retrieval on the web, 29–32.
jects. The next major running theme on OGD research is about issues Berners-Lee, T. (2006). Linked data-design issues. http://www.w3.org/DesignIssues/
LinkedData.html.
relating to citizen engagements. Unlike FOI, the major themes identi- Birkinshaw, P. (2010). Freedom of information and its impact in the United Kingdom.
fied in OGD research do not closely reflect much of its touted principles Government Information Quarterly, 27(4), 312–321. http://dx.doi.org/10.1016/j.giq.
and key operational terms. This is because the topic label Transparency/ 2010.06.006.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. Journal of
Collaboration/Participation which are terms which have mostly occa- Machine Learning Research, 3(January), 993–1022.
sioned the OGD campaign, conspicuously had very few proportion of Blei, D. M. (2012). Probabilistic topic models. Communications of the ACM, 55(4), 77–84.
the topics. This reinforces the point that much of the focus on OGD http://dx.doi.org/10.1145/2133806.2133826.
Camaj, L. (2016). From ‘window dressing’to ‘door openers’? Freedom of Information
research have been on technological issues relating to data access and legislation, public demand, and state compliance in South East Europe. Government
management of public data. Information Quarterly, 33(2), 346–357. http://dx.doi.org/10.1016/j.giq.2016.03.001.
In respect of research question 2, the results by the topic modelling Carrasco, C., & Sobrepere, X. (2015). Open government data an assessment of the spanish
particularly revealed some other di erences between freedom of in- municipal situation. Social Science Computer Review, 33(5), 631–644. http://dx.doi.ff org/10.1177/0894439314560678.
formation and open government data that have not yet been mentioned Chang, J., Gerrish, S., Wang, C., Boyd-Graber, J. L., & Blei, D. M. (2009). Reading tea
in the literature like those addressed by Geiger and von Lucke (2012), leaves: How humans interpret topic models. Proceedings in Advances in Neural
Janssen (2012) and Ubaldi (2013). One of such difference identified in
this study, is the issue of cost to an individual accessing a public data.
Though the focus in the two campaigns is to progressively make public 7 https://www.data.gov/open-gov/.
671
E. Afful-Dadzie, A. Afful-Dadzie International Journal of Information Management 37 (2017) 664–672
Information Processing Systems, 288–296. Dirichlet allocation and AdaBoost. Proceedings of 2012 IEEE conference on intelligence
Charalabidis, Y., Alexopoulos, C., & Loukis, E. (2016). A taxonomy of open government and security informatics, 102–107.
data research areas and topics. Journal of Organizational Computing and Electronic Relly, J. E., & Schwalbe, C. B. (2016). How business lobby networks shaped the US
Commerce, 26(1–2), 41–63. http://dx.doi.org/10.1080/10919392.2015.1124720. Freedom of Information Act: An examination of 60years of congressional testimony.
Chignard, S. (2013). A brief history of open data. Paris Tech Review29. Government Information Quarterly, 33(3), 404–416. http://dx.doi.org/10.1016/j.giq.
DiMaggio, P., Nag, M., & Blei, D. (2013). Exploiting affinities between topic modeling and 2016.05.002.
the sociological perspective on culture: Application to newspaper coverage of US Roberts, A. (2010). A great and revolutionary law? The first four years of India’s Right to
government arts funding. Poetics, 41(6), 570–606. http://dx.doi.org/10.1016/j. Information Act. Public Administration Review, 70(6), 925–933. http://dx.doi.org/10.
poetic.2013.08.004. 1111/j.1540-6210.2010.02224.x.
Donnelly, J. (2013). Universal human rights in theory and practice. Cornell University Press. Rostron, A. (2013). The mugshot industry: Freedom of speech, rights of publicity, and the
Downey, D., Broadhead, M., & Etzioni, O. (2007). Locating complex named entities in controversy sparked by an unusual new type of business. Washington University Law
web text. Proceedings of International Joint Conference on Artificial Intelligence (IJCAI), Review, 90(4).
2733–2739. Schartum, D. W. (1998). Access to government-held information: Challenges and possi-
Evans, A. M., & Campos, A. (2013). Open government initiatives: Challenges of citizen bilities. The Journal of Information, Law and Technology. http://www2.warwick.ac.uk/
participation. Journal of Policy Analysis and Management, 32(1), 172–185. http://dx. fac/soc/law/elj/jilt/1998_1/schartum/.
doi.org/10.1002/pam.21651. Shadbolt, N., O'Hara, K., Berners-Lee, T., Gibbins, N., Glaser, H., & Hall, W. (2012).
Foerstel, H. N. (1999). Freedom of information and the right to know: The origins and ap- Linked open government data: Lessons from data.gov.uk. IEEE Intelligent Systems,
plications of the Freedom of Information Act. Greenwood Publishing Group. 27(3), 16–24.
Frey, L. (2014). Open government partnership four-year strategy 2015–2018. http://www. Shepherd, E., Stevenson, A., & Flinn, A. (2009). The impact of freedom of information on
opengovpartnership.org/sites/default/files/attachments/4YearAP-Online.pdf. records management and record use in local government: A literature review. Journal
Geiger, C. P., & von Lucke, J. (2012). Open government and (linked)(open)(government) of the Society of Archivists, 30(2), 227–248. http://dx.doi.org/10.1080/
(data). JeDEM-eJournal of eDemocracy and Open Government, 4(2), 265–278. 00379810903445000.
Geissbuhler, A., Safran, C., Buchan, I., Bellazzi, R., Labkoff, S., Eilenberg, K., et al. (2013). Tauberer, J. (2012). History of the movement. Open government data: The bookhttps://
Trustworthy reuse of health data: A transnational perspective. International Journal of opengovdata.io/2014/history-the-movement/.
Medical Informatics, 82(1), 1–9. http://dx.doi.org/10.1016/j.ijmedinf.2012.11.003. Tinati, R., Carr, L., Halford, S., & Pope, C. (2012). Exploring the impact of adopting open
Gigler, B. S., Custer, S., & Rahemtulla, H. (2011). Realizing the vision of open government data in the UK government. Digital futures. Aberdeen, GB: Web & Internet Science 3pp.
data: Opportunities, challenges, and pitfalls. World Bank. US. Senate (2007). Open government: Reinvigorating the freedom of information act: Hearing
Halstuk, M. E., & Chamberlin, B. F. (2006). The Freedom of Information Act 1966–2006: before the committee on the judiciary, United States senate, one hundred tenth congress,
A retrospective on the rise of privacy protection over the public interest in knowing first session, march 14, 2007. Washington: U.S. G.P.O. DIANE Publishing. https://
what the government’s up to. Communication Law and Policy, 11(4), 511–564. http:// www.gpo.gov/fdsys/pkg/CHRG-110shrg35801/pdf/CHRG-110shrg35801.pdf.
dx.doi.org/10.1207/s15326926clp1104_3. Ubaldi, B. (2013). Open government data: Towards empirical analysis of open government
Power, voting, and voting power. In M. J. Holler (Ed.), Springer Science & Business data initiatives. Paris, France: OECD.
Media. Van Dooren, W., Bouckaert, G., & Halligan, J. (2015). Performance management in the
Hubbard, P. (2008). China’s regulations on open government information: Challenges of public sector. Routledge.
nationwide policy implementation. Open Government: a Journal on Freedom of Veljković, N., Bogdanović-Dinić, S., & Stoimenov, L. (2014). Benchmarking open gov-
Information, 4(1), 1–34. ernment: An open data perspective. Government Information Quarterly, 31(2),
Janssen, M., Charalabidis, Y., & Zuiderwijk, A. (2012). Benefits, adoption barriers and 278–290. http://dx.doi.org/10.1016/j.giq.2013.10.011.
myths of open data and open government. Information Systems Management, 29(4), Wang, C., & Blei, D. M. (2011). Collaborative topic modeling for recommending scientific
258–268. articles. Proceedings of the 17th ACM SIGKDD international conference on Knowledge
Janssen, K. (2012). Open government data and the right to information: Opportunities discovery and data mining (pp. 448–456).
and obstacles. The Journal of Community Informatics, 8(2). Wang, H. J., & Lo, J. (2016). Adoption of open government data among government
Jetzek, T., Avital, M., & Bjørn-Andersen, N. (2012). The value of open government data: A agencies. Government Information Quarterly, 33(1), 80–88. http://dx.doi.org/10.
strategic analysis framework. Proceedings of the 2012 pre-ICIS workshop. 1016/j.giq.2015.11.004.
Jiang, H., Qiang, M., & Lin, P. (2016). A topic modeling based bibliometric exploration of Wang, Q., & Waltman, L. (2016). Large-scale analysis of the accuracy of the journal
hydropower research. Renewable and Sustainable Energy Reviews, 57, 226–237. classification systems of Web of Science and Scopus. Journal of Informetrics, 10(2),
Krishnamurthy, R., & Awazu, Y. (2016). Liberating data for public value: The case of data. 347–364. http://dx.doi.org/10.1016/j.joi.2016.02.003.
gov. International Journal of Information Management, 36(4), 668–672. Wei, X., & Croft, W. B. (2006). LDA-based document models for ad-hoc retrieval.
Lathrop, D., & Ruma, L. (2010). Open government: Collaboration, transparency, and parti- Proceedings of the 29th annual international ACM SIGIR conference on research and
cipation in practice. O’Reilly Media, Inc. development in information retrieval, 178–185.
Lau, R. R., Patel, P., Fahmy, D. F., & Kaufman, R. R. (2014). Correct voting across thirty- Weisberg, H. F., & Nawara, S. P. (2010). How sophistication affected the 2000 pre-
three democracies: A preliminary analysis. British Journal of Political Science, 44(02), sidential vote: Traditional sophistication measures versus conceptualization. Political
239–259. Behavior, 32(4), 547–565. http://www.jstor.org/stable/40960954.
Mann, G. S., Mimno, D., & McCallum, A. (2006). Bibliometric impact measures leveraging Whitmore, A. (2012). Extracting knowledge from US department of defense freedom of
topic analysis. Proceedings of the 6th ACM/IEEE-CS joint conference on digital libraries, information act requests with social media. Government Information Quarterly, 29(2),
65–74. 151–157.
Maskeri, G., Sarkar, S., & Heafield, K. (2008). Mining business topics in source code using Witten, I. H. (2005). Text mining. Practical handbook of internet computing, 14-1. CRC-
latent dirichlet allocation. Proceedings of the 1st India software engineering conference computer-and-information-science-series/book-series/CHCOMINFSCI. Chapman-Hall.
(pp. 113–120). Xing, D., & Girolami, M. (2007). Employing latent Dirichlet allocation for fraud detection
Maude, F. (2012). Open data white paper-unleashing the potential. the stationary office limited in telecommunications. Pattern Recognition Letters, 28(13), 1727–1734. http://dx.doi.
on behalf of HM government. London, United Kingdom: Cabinet Office. org/10.1016/j.patrec.2007.04.015.
Mendel, T. (2008). Freedom of information: A comparative legal survey. Paris: Unesco. Yau, C. K., Porter, A., Newman, N., & Suominen, A. (2014). Clustering scientific docu-
Novais, T., de Albuquerque, J. P., & da Silva Craveiro, G. (2013). An account of research on ments with topic modeling. Scientometrics, 100(3), 767–786. http://dx.doi.org/10.
open government data (2007–2012): A systematic literature review. EGOV/ePart ongoing 1007/s11192-014-1321-8.
research (pp. 76–83). Zeleti, F. A., Ojo, A., & Curry, E. (2016). Exploring the economic value of open govern-
Owen, G. T. (2014). Qualitative methods in higher education policy analysis: Using in- ment data. Government Information Quarterly, 33(3), 535–551. http://dx.doi.org/10.
terviews and document analysis. The Qualitative Report, 19(26), 1. 1016/j.giq.2016.01.008.
Ríos, S. A., Aguilera, F., Bustos, F., Omitola, T., & Shadbolt, N. (2013). Leveraging social Zuiderwijk, A., & Janssen, M. (2014). The negative effects of open government data-
network analysis with topic models and the Semantic Web (extended). Web investigating the dark side of open data. Proceedings of the 15th ACM annual inter-
Intelligence and Agent Systems: An International Journal, 11(4), 303–314. national conference on digital government research, 147–152.
Ramanathan, V., & Wechsler, H. (2012). Phishing Website detection using latent
672
View publication stats