Social Media Domain Analysis (SoMeDoA)
A Pharmaceutical Study
David Bell and Sara Robaty Shirzad
Brunel University, Information Systems and Computing, London, U.K.
Keywords: Social Media Network (SMN), e-Marketplace, Business to Business Environment.
Abstract: Social media data is increasingly becoming a valuable asset for marketing teams, and businesses are
regularly coming up with new and innovative ways to make use of this data. A social media network (SMN)
is able to connect enterprises with their customers, partners and even competitors. Public trading and
relations-oriented structures of social media networks (SMN) have encouraged organizations to engage
more actively with other transactional partners. Organizations are seeking to tap into the relationship
development potential these websites offer, especially the network effect of each individuals or
organisations social graph. It is recognized that these relationships (when utilised) are able to create value
for network participants. This paper discusses SMN tools and outlines a methodology and procedure that
supports the identification of domain specific networks within a global business-to-business environment.
Research is carried out using SMN data about firms in the pharmaceutical industry. We use our own
methodology to uncover market participants, linkages and prominent issues that may help new firms to
position themselves effectively in a new marketplace. SMNs provide a considerable source of information
and new methods are required to fully leverage their potential value. This paper explores how SMNs can be
used as an effective source of business intelligence by analysing a popular SMN platform.
1 INTRODUCTION
As markets increase globally, smaller firms, whether
they want to be or not, find themselves to be part of
social networks (Pitt et al., 2006). Consequently,
increased connectivity to customers results in
increased competition with rivals from around the
world. Being in the social networks could even be an
opportunity for them to survive and compete with
larger counterparts (Copp and Ivy, 2002; Masurel
and Janszen, 1998; Lipparini and Sobrero, 1994). In
September 2012, CNBC, a recognized leading
website in business news, reports that about 15.2
million site members in LinkedIn (one of the social
network websites for professionals) or about 8.7 per
cent of the site's 175 million members worldwide are
small-business professionals. The strategic choice of
new members of social networks is simple: how to
understand these social networks in order to get
them work for their own business. There is lack of
empirical studies on social networks that try to
answer questions around knowledge of discrete
business networks and the advent of internet
provides a unique opportunity to study these
business interrelationships. The Internet itself could
be considered as the largest social network.
Moreover, the internet is becoming the most popular
vehicle for business to business (B2B) commerce. In
this article, we examine the social network facing
small to medium firms (SMEs) in the
pharmaceutical industries. To achieve this aim, we
analyse activity of a number of large organizations
in the same domain to understand how large
companies are cultivating relationships in SMN’s so
that small-medium organizations could explore their
relationships to customers and counterparts.
We use the framework to recognize prominence
that helps the new firms to position themselves
effectively in a social network to leverage
considerable value. This paper aims to improve the
understanding on how SMNs can be used as a
reliable source by analysing the temporal on Twitter
1
activities. Twitter was chosen for the popularity,
according to a recent study (Skeels and Grudin,
2009) one third of employees in enterprises are in
Twitter company networks. Professionally oriented
1
http://www.twitter.com
561
Bell D. and Robaty Shirzad S..
Social Media Domain Analysis (SoMeDoA) - A Pharmaceutical Study.
DOI: 10.5220/0004499105610570
In Proceedings of the 9th International Conference on Web Information Systems and Technologies (STDIS-2013), pages 561-570
ISBN: 978-989-8565-54-9
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
structures within Twitter make it popular among
organizations. Although the methodology outlined
from the study has only involved one domain (albeit
a large one), it is argued that the richness of the
information provided by users from different
backgrounds will provide generalizable outcomes to
a range of scenarios. The paper is structured as
follows: First, we provide a brief overview of social
media commerce. Next we describe a methodology
for exploring the data. Third, we describe the results
of the studies. After identifying the obvious
limitations of the research, we conclude and discuss
future opportunities for new entrants to the e-
Marketplace.
2 SOCIAL MEDIA COMMERCE:
A PLATFORM FOR FINDING
CONNECTIONS
With the large amount of information potentially
available to organizations, the internet constitutes an
important platform for information exchange
between the consumers and industry suppliers,
intermediaries, as well as organizations which are
not have experience of being in the e-marketplaces.
Different technology interfaces such as search
engines, intermediaries facilitate the marketing
information exchange between online organizations.
Social media, which enables interaction among
virtual organizations, has emerged as an integral
element of communication. Ellison (2007) defines
social media sites as Web-based services that allow
individuals to (1) construct a public or semi-public
profile within a bounded system, (2) articulate a list
of other users with whom they share a connection,
and (3) view and traverse their list of connections
and those made by others within the system. Noted
by Harris (2009), there are hundreds of different
social media platforms such as text messaging,
shared photo, social network, wikis, and discussion
group. According to Alexa (2012), a Web
information company that provides website traffic
rankings, top 5 global social media websites by late
2012 that have significant presence on enterprises
are: (1) Facebook, (2) Twitter, (3) LinkedIn, (4)
MySpace and (5) Google Plus+. Some of these
websites are more likely accessed by youths with IM
experience such as Facebook. Some of them like
LinkedIn and Twitter target professional use from
the start.
In general, social media websites are beneficial
and valuable for the network participants in that they
promote activities and the use of resources
(Michaelidou, Siamagka and Christodoulides, 2011;
Walter et al., 1997). A study published on February
2010 by the Small Business Success Index (SBSI)
indicates that 75% of the surveyed small businesses
in the USA have already created a company page on
a social networking site. A number of researchers
are focused on social media theory to study the
social networks of firms in B2B marketplaces (e.g.
(Michaelidou, Siamagka and Christodoulides, 2011;
Björkman and Kock, 1995)). Further, number of
studies on small firms argues that social networks
are important for the survival of small firms and
critical in competing with large businesses (Pitt et
al., 2006; Copp and Ivy, 2002). Indeed social media
websites are suited for collecting
information/feedback from customers, developing
relationships between customers through interaction.
Some of the interactions from organizations and
supporters in social media websites summarized by
Carrera et al. (2008) are: posting links to external
news item about the organization or its causes; using
messaging broad or discussion wall to post-
announcements and answer questions.
Adopting new channels of technology such as
social media websites may not be attractive to many
companies. Many researches stress that many
organizations have been slow to adopt new
technologies due to perceived barriers such as lack
of money, time and training, negative views about
usefulness, as well as unfamiliarity with technology
(Buehrer, Senecal and Bolman Pullins, 2005;
Venkatesh and Davis, 2000). Additionally,
Frambach and Schillewaert (2002) argue that
organization size is also important in the adoption
decision. They further suggest that smaller
organizations are more innovative and are therefore
expected to be more receptive to new changes in
technologies. Copp (2002) and Pitt (2006) highlight
that social network websites are important for the
survival of small firms and critical in competing
against large businesses. Increasing the number of
socially active organizations change the strategic
view of other partners to consider social networks as
important arena to consider for competing in virtual
marketing world. Figure 1, illustrates a conceptual
framework of social media commerce that an
organization seeks to participate in B2B marketplace
and getting information about that marketplace
through social media networks. Social media
networks in this framework are a knowledge source
for newly entering organizations.
One of the advantages of being active in social
media websites is notable in their design in relation
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
562
Figure 1: Social Media Commerce.
to both time and space (De Longueville, Smith and
Luraschi, 2009). For instance, tweets on Twitter are
organised in timelines (i.e. series of tweets sorted
and displayed in reverse chronological order) (De
Longueville, Smith and Luraschi, 2009). On the
other hand, Social media websites are structured
architecturally so as to communicate a space (e.g.
geo-location in twitter) that is publicly accessible.
Thus, private and public boundaries are employed to
situate the network geographically (Papacharissi,
2009). Longueville et al. (2011) explored the role of
Twitter as source of spatio-temporal information for
retrieving, validating and filtering spatio-temporally
referenced images from Flickr within the context of
flooding in the UK to advance existing capabilities
for monitoring natural hazards. They emphasized
that graphical representation of place names as
spatial references (e.g., town, county, etc.), resulted
in space accuracy from a gazetteer viewpoint. In
contrast, temporal referencing is the time-stamp
when something posted on social media's. Lee et al.
(2011) (Lee et al., 2011) plotted information from
twitter on earthquakes on a world map. They used
city names to define positions of tweets.
Although, the body of scientific literature about
social media websites are growing (as an example,
the online scientific literature database Scopus.com
provides 88 results for 2009, 123 results for 2010,
194 results for 2011 and 259 results for 2012 while
searching for the keyword social media websites), its
potential for spatio-temporal information has still to
be fully assessed. Most of the research is focused on
social aspects by analysing users satisfactions (Java
et al., 2009), interactions of users with social
websites (Huberman, Romero and Wu, 2008) or
adoption (Hughes and Palen, 2009). Pharmaceutical
e-marketplace adoption has been volatile over the
past decade (Shirzad, 2013). This paper aims to
provide further exploration of the role of social
media websites such to identify the motiving factors
for pharmaceutical organizations to participate in e-
marketplaces and to consider the valuable temporal
and geospatial components they can contain. Our
research question requires us to advance spatio-
temporal analysis methods in order that suitable
knowledge can be extracted from social media
networks for newly entering organizations to the
market. In the remainder of this paper, possible
methods are described, using a real-life datasets as
an illustration. This work can serve as a theoretical
base for future works aiming at performing web-
based event, behaviour and strategy detection.
3 SOCIAL MEDIA DOMAIN
ANALYSIS - SoMeDoA
APPROACH
The temporal model considers the visit activity of
people to specific times (including intervals) (Yoo
and Hwang, 2008). The idea of harvesting temporal
information from the Web has been of interest in
recent years (De Longueville, Smith and Luraschi,
2009). Li et al. (2005) proposed a probabilistic
model to detect retrospective news events. They
explained the generation of “four Ws” - who
(persons), when (time), where (locations) and what
(keywords), from each news article. However, they
considered time and location for discovering the
reoccurring peaks in events. Mei et al. (2006)
produced a model for spatio-temporal analysis for
SocialMediaDomainAnalysis(SoMeDoA)-APharmaceuticalStudy
563
Table 1: SoMeDoA Research Framework.
Phase Description Resulting output
Data
Selection
Social media Web sites are selected as suitable
sources for the domain of study.
List of social media platforms and associated search
terms.
Data
Gathering
Data gathering tools are selected and run
against the selected Social Media sites.
List of software tools.
Generated data files.
Temporal
Separation
Public information, news and communications
are extracted in order to determine the public
activities of organisations (with associated
timelines)
DateTime lists files for each organisation.
Temporal
Coding
Further analysis of temporal data in order to
uncover topics of importance (with timeline)
Keyword lists and domain ontology
DataTime Data lists for each keyword, code or
category
weblog data. In contrast to previous work, we apply
the temporal model to describe organizations
activities on Twitter with a more explorative
motivation.
Our high level research framework is presented
Table 1 and titled Social Media Domain Analysis
(SoMeDoA). It involves core elements of social
media data gathering and data analysis (including
Grounded Theory approaches). Data from specific
social media web sites is extracted using domain
specific search terms that each target particular
organisations temporal data sets. The generated data
files are then analysed using a mix of visualisation
and analytical tools.
Our research approach “in action” consisting of
two main stages: temporal separation and temporal
coding, converting data from Twitter into temporal
aspects of organizations information. Twitter is
selected in order to effectively detect real-time
activity of organizations within a domain (the
“what” and “when”).
We use the name of the organizations as a query
term to get the tweets, they (or others) publish.
Subsequently, Tweetcatcher2 (an application
developed on the MATCH project at Brunel
University) is used to retrieve tweets and related
data such as published date, user ID, tweet, number
of following, number of followers, time zone,
number of users tweets, retweet count, expanded
links and sentimental analysis (Sentiment analysis
assigns scores indicating positive, neutral and
negative opinion to each distinct entity in the text
(Pak and Paroubek, 2010). The temporal separation
and coding analysis activities are developed for
handling Twitter message streams, and categorized
them with respect to the number of tweets published
and frequency of occurrences in the selected
timeslots. Temporal separation analysis was carried
out in Microsoft Excel 2010. The dataset is
visualized based on the time and number of tweets
generated. The second part of Twitter analysis was
temporally coding. The approach taken for the
temporal coding analysis of content was based on
the grounded theory method (GTM). GTM is the
process of generating a theory from collected data
(Glaser and Strauss, 1967). We used Nvivo9 (as
software to support grounded theory techniques) in
order to analyse tweet data. It was used for content
analysis: 1) Storage and categorizing datasets, 2)
conducting searches for further analysis in order to
generate reports about frequency of word
occurrences and associated categorisations and 3)
creation of categories through computer-assisted
coding. For example, a financial innovation category
was created with associations to acquisition, finance
and investor. Tweet frequency was used as a guide
to category and sub-category importance. Both
temporal separation and coding continue on with
sentimental analysis with respect to time and
wording. The data used in this study were collected
in November 2012. A total of 54365 tweets were
captured posted by selected organizations. Social
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
564
media data (including data interfaces) offer structure
to data not found with traditional Web mining. Field
descriptors in the Web sites’ data interface or
annotation (e.g. hashtags) in the data both offer
opportunities for improved analyses.
4 EXPERIMENTS
AND RESULTS
The pharmaceutical domain has actively used e-
marketplaces over the past decade with platform
volatility in line with evolutions on the Web
(Shirzad, 2013). Further exploration is required into
the role of social media websites can play in
identifying the motiving factors for pharmaceutical
organizations to participate in e-marketplaces. The
first step is to fully understand the business network
in which they operate and to consider the value of
temporal components that are accessible.
4.1 Twitter Temporal Separation
Temporal analysis deals with time components
(Lauw et al., 2005). So far we have been monitor
tweets posted daily from 11
th
to 29
th
of November
2012. As mentioned before, the top five
pharmaceutical organizations were selected from
Fortune Global which is an official website for
ranking pharmaceutical organizations. 54,365 tweets
were posted and subsequently downloaded for
analysis. In order to calculate the proportion of
organization activities on Twitter, we divide the
dataset into three weekly time slots. Later on we
decide to analyse Twitter activity on each
Wednesday of weeks in November.
4.1.1 Tweets per Week
The first step was to analyse the overall number of
tweets and how the numbers of tweets published are
varying week by week. Figure 2 shows a graph with
the total number of tweets between 14.11.2012 and
28.11.2012. The first and last tweet mentioning in all
three time slots were published at 00:00 and 23:59
respectively. The columns are positioned over a
label that represents the date and time that tweets
posted. The height of the column indicates the
number of tweets that the chosen organizations
(under analysis) posted, defined by the column label.
The highest rise in number of tweets was on 21
th
of
November (Figure 2-B) where the total number of
38,272,550 tweets was published. Whereas the
number of tweets on 14
th
(Figure 2-A) and 28
th
(Figure 2-C) were 33,992,110 and 10,001,050
respectively. The other interesting issue in figure 2 is
that the number of Tweets per minute on figure 2-C
is more than the other time slots. For example on
28
th
of November, the peak tweets content were
about online buying medicine from Roche (one of
the chosen organizations). Most of the tweets were
about new way of doing the online buying the
medicines through the new portal. Also the highest
number of re-tweet is on the same organization. The
motivation behind this is investigating how useful is
the sentiment lexicon developed for tweets in these
three time slots. Peaks can be investigated for
additional knowledge. In one example, a peak
includes tweets about Warren Buffet’s Berkshire
Hathaway, whose sale of Johnson & Johnson shares
are reported in the mainstream media.
4.1.2 Sentimental Average per Week
To assign numerical scores to sentiments of an
individual sentence, we use the SentiStrength 7 tool
developed by Thelwall et al. (2010) (Thelwall et al.,
2010) and used by Brunel’s Tweetcatcher. This tool
simultaneously assigns both a positive and a
negative score to bits of English text, the idea being
that users can express both types of sentiments at the
same time, such as in “I love you, but I also hate
you” (Kucuktunc et al., 2012). Positive sentiment
strength scores range from +1 (not positive) to +5
(extremely positive) and similarly, negative
sentiment strength scores range from −1 to −5
(Kucuktunc et al., 2012). The final positive
sentiment strength for a bit of text is then computed
by taking the maximum score among all individual
positive scores. The negative sentiment strength is
similarly calculated. Figure 3 shows the distribution
of sentiment.
The vast majority of sentences are assigned a
neutral +1/-1 sentiment score. Slightly negative
(+1/-2) and slightly positive (+2/-1) scores are also
common. Table 2 shows the percentage of each
score in each time slot.
Table 2 clearly shows that the percentages of
positive and negative tweet sentiments are much
higher in first and last weeks of the month. The
weekly sentiment analysis doesn’t indicate the actual
content of the tweets (e.g. Berkshire Hathaway’s
share sale mentioned earlier) beyond weekly
sentiment analysis. Therefore temporal coding was
conducted. This leads us to investigate more about
tweets content by counting a word frequency for all
the dataset and undertake sentiment analysis on
frequently occurring words.
SocialMediaDomainAnalysis(SoMeDoA)-APharmaceuticalStudy
565
4.2 Twitter Temporal Coding
4.2.1 Tweet per Words
In the first instance, frequently used words or topics
need to be identified in order to get a picture of the
actual tweet content. Data needs to be subject to
careful scrutiny and interpretation, which is largely
achieved through a coding process. The approach
taken for analysing content makes use of the
grounded theory method (GTM). GTM is the
process of generating a theory from collected data
(Glaser and Strauss, 1967). The process was
conducted by counting the word frequency for the
dataset of tweets using Nvivo9. The most frequent
words were “http” followed by other parts of URLs
that appeared in most of the tweets (which should be
discounted). After excluding articles and other terms
that did not provide meaningful context, table 3
shows the most frequent words.
As table 3 presents, the most frequent words are
Roche followed by Johnson that appears in the most
of the tweets. At this level we can get a general
impression of key players and typical work
associations (e.g. news, sales). Sentiment analysis on
the most frequent words will help us to understand
more on positive or negative tweets about those
words over time.
4.2.2 Sentimental Average per Word
The pharmaceutical organizations are unsurprisingly
most occurring words (as they are the search terms
in question). Therefore, we first decide to do
sentiment analysis on tweets to see how many
positive and negative tweets published on each
organization name. Table 4 presents the overall view
Figure 2: User Tweets per week.
Figure 3: Tweets-Sentimental Average.
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
566
Table 2: The distribution of sentiment scores.
Time Slot Senti-Positive Senti-Neutral Senti-Negative
14
th
November 2012 32% 47% 21%
21
th
November 2012 25% 53% 22%
28
th
November 2012 27% 48% 25%
Table 3: Frequent words in tweets.
Frequent words in tweets
Between 3000-4000 occurrences Roche
Between 2000-3000 occurrences Johnson & Johnson
Between 1000-2000 occurrences Pfizer, Novartis
Between 0-1000 occurrences Glaxo, GSK, innovations, news, marketing, yahoo, finance, healthcare,
acquisition, city, advertising, business, development, manufacturing,
products, research, investors, competition, consumer, customer, email,
auction, collaboration, communication, contract, social network,
supplier, Facebook, Google, government, outsourcing, technology, e-
pharma, distributors, economics, twitter
on the number of positive and negative tweets
published and the average frequency of occurrence
for each organisation.
The same sentiment analysis is carried out on
these terms to determine their respective time lines
of sentiment. The timelines (Figures 2) can then be
generated for both the companies and the categories.
Tweet frequency was used as a guide to category
and sub-category importance. In order to infer some
of the socially based relationships, the content
analysis is carrying out. The knowledge structure
derived from tweets (ontology) is presented in next
section.
4.2.3 Ontology and Concept Network
Ontology network is relating ontologies on the basis
of explicit import relationships or implicit similarity.
Concept network is relating concepts on the basis of
explicit ontological relationships or implicit
similarity. As mentioned in section 3, GTM taken in
order to analyse the tweets content. Figure 4 presents
the process of storing, categorizing datasets.
The coding of content resulted in a number of
categories and subcategories, including:
Technology, Finance, Innovation, Suppliers,
Government, Healthcare, Investors. For example, a
financial innovation category was created with
associations to acquisition, finance and investor.
Subsequently, we have extracted the above
mentioned ontology by folding the graph using the
Protégé 4.2 OntoGraf. The same sentiment can then
be report by key codes and categories.
SocialMediaDomainAnalysis(SoMeDoA)-APharmaceuticalStudy
567
Table 1: Senti-average per frequent Word (Organisations).
Name #Senti-Pos. #Senti-Neg. Senti-Pos.
percentage
Senti-Neg.
percentage
Senti-Pos. Average Senti-Neg.
Average
Roche 5304 4444 56% 48% 1.33195 -1.56133
Johnson
&
Johnson
2134 848 23% 9% 1.363636 -1.35142
Pfizer 1058 2653 11% 28% 1.413043 -1.39224
Novartis 747 1205 8% 13% 1.570281 -1.62905
GSK 166 163 2% 2% 1.150602 -1.20245
Figure 1: Tweets coding and categorizing.
5 CONCLUDING DISCUSSION
Before attempting to interpret and make sense of the
findings, a caveat should be highlighted with regards
to the dataset being analysed. The exploratory nature
of this study and its focus on methodology meant
that it was not possible to examine in further depth
the actual content of the over 54,365 tweets
captured. A more elaborate classification of tweet
patterns is an issue of further research. Bearing this
limitation in mind, it seems that pharmaceutical
organizations are at the process of building an
extensive LinkedIn and Twitter networks that gives
them access to a diverse group of stakeholders.
These same networks are freely available and
provide a useful source of business intelligence –
especially for small or new entrant organisations.
This is indicated not only by the promising number
of tweets per account, but also by the frequency of
occurrence of the words, as well as the sentiment
analysis across accounts, which are usually quite
influential. This diversity confirms that, beyond
interactivity with organizations, LinkedIn and
Twitter accounts indeed attempt to satisfy quite
complicated information needs for organizations in
terms of activity on social networks.
The findings of our study outline certain mixed
conclusions about this ability. Firstly, as far as the
content produced by the accounts is concerned, there
are two main observations. The first is that the
content is obviously temporal as indicated by the
popular timeslot and keywords identified in the
analysis. Temporal separation (weekly based)
analysis on the content of Twitter dataset was
carrying on in order to compare the number of
tweets published in each time slot (figure 2). The
result over the three snapshot times show the rise in
number of tweets on second week of the month (21
November). Whereas the number of tweets per
minutes in the last timeslot (28 November) is more
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
568
Figure 5: Content network.
transparent. Additionally our temporal separation
analysis followed by sentiment information analysis
on the time-stamps. We demonstrated the average of
number of positive and negative tweets per each
timeslot (figure 3, table 2). Sentiment analysis
experiment shows the vast majority of tweets are
assigned to neutral. The analysis shows tweets are
slightly more positive in November 14
th
whereas
towards end of the month the sentiments are more
negative. Thirdly, further analysis was carried out on
tweet content. Analysis was carried out using our
own novel temporal coding technique drawing in
part from the grounded theory method. In temporal
coding, initially we count the number of word
occurrences using Nvivo9. Word frequency analysis
revealed 5 major words of those who tweet: Roche,
Johnson &Johnson, Pfizer, Novartis and GSK. This
categorization proved important in better
understanding the type of contents those re-using
such as content will be faced with. In particular we
highlighted that unsurprisingly, few tweets are
recognized as rich in content and therefore valuable
in their own right. Alternatively, in aggregated form
(by organisational, time, word or coded grouping) an
interesting picture can be uncovered. This is because
twitter users provide a mix of information which
can’t be easily distinguished by automatic means.
Sentiment analysis (coupled with Grounded Theory
categorisation) offers a number of opportunities to
better understand the wider business networks and
their language. It can be seen that Twitter offers a
promising starting point for crawlers to collect
related data, where time and location matter – a
domain picture.
This paper has analysed datasets from a popular
social media network. Such a subtle difference in
social networks leads us to think more about
semantic integration to achieve interoperability
among SMNs and ultimately content integration
facilities on the Web. Semantic integration can
provide an enhanced view of individual or
organizations activities in distributed social
networks. Therefore more intuitive semantic
methods are required for presenting and navigating
around SMN data. On the other hand, analysing data
published on social network provide a unique
opportunity for an SME or new entrant to the market
to observe the dynamics of community development
of large organizations as the data is easy, cheap and
accessible.
In this study, the research approach and social
networking analysis provided the opportunity to
investigate the pharmaceutical organizations Twitter
accounts beyond sampling a particular set of tweets
or focusing only on a small subset of those accounts.
Such an investigation could not be possible by
applying standard research methods, which are not
able to follow the pace of Internet change as
organisations develop their online presence (Karpf,
2012). Therefore, research methods need to be
informed accordingly so that the complex
interactions being formed on social networks can be
adequately understood. The Social Media Domain
Analysis (SoMeDoA) framework is a key
contribution of this paper and attempts to motivate
more rigorous and integrative approaches to social
media data analysis. The framework provides an
effective approach to both selecting (and integrating)
social media platforms and subsequent analysis of
the data they hold. Temporal categorisation of data
(including the addition of key domain concepts)
before frequency and sentiment analysis provides an
effective means of researching a domain of interest,
marketplace or industry. The SoMeDoA approach
SocialMediaDomainAnalysis(SoMeDoA)-APharmaceuticalStudy
569
can be practically applied to further domains of
study (or lines of inquiry), providing a flexibly
method for mining new insight from readily
available data.
ACKNOWLEDGEMENTS
The authors acknowledge support of this work
through the MATCH Programme (UK Engineering
and Physical Sciences Research Council grants
numbers GR/S29874/01, EP/F063822/1 and
EP/G012393/1).
REFERENCES
Björkman, I. and Kock, S. (1995) "Social relationships
and business networks: The case of western companies
in China", International Business Review, vol. 4, no. 4,
pp. 519-535.
Buehrer, R. E., Senecal, S. and Bolman Pullins, E. (2005)
"Sales force technology usage—reasons, barriers, and
support: An exploratory investigation", Industrial
Marketing Management, vol. 34, no. 4, pp. 389-398.
Copp, C. B. and Ivy, R. L. (2002) "Networking Trends of
Small Tourism Businesses in PostSocialist
Slovakia", Journal of Small Business Management,
vol. 39, no. 4, pp. 345-353.
De Longueville, B., Smith, R. S. and Luraschi, G. (2009)
"Omg, from here, i can see the flames!: a use case of
mining location based social networks to acquire
spatio-temporal data on forest fires", Proceedings of
the 2009 International Workshop on Location Based
Social NetworksACM, , pp. 73.
Glaser, B. G. and Strauss, A. L. (1967) The discovery of
grounded theory: Strategies for qualitative research,
AldineTransaction.
Huberman, B., Romero, D. and Wu, F. (2008) "Social
networks that matter: Twitter under the microscope",
Available at SSRN 1313405, .
Hughes, A. L. and Palen, L. (2009) "Twitter adoption and
use in mass convergence and emergency events",
International Journal of Emergency Management, vol.
6, no. 3, pp. 248-260.
Java, A., Song, X., Finin, T. and Tseng, B. (2009) "Why
we twitter: An analysis of a microblogging
community", Advances in Web Mining and Web
Usage Analysis, , pp. 118-138.
Karpf, D. (2012) "Social Science Research Methods in
Internet Time", Information, Communication &
Society, vol. 15, no. 5, pp. 639-661.
Kucuktunc, O., Cambazoglu, B.B., Weber, I. and
Ferhatosmanoglu, H. (2012) "A large-scale sentiment
analysis for Yahoo! answers", Proceedings of the fifth
ACM international conference on Web search and
data miningACM, , pp. 633.
Lauw, H. W., Lim, E.P., Pang, H. and Tan, T. T. (2005)
"Social network discovery by mining spatio-temporal
events", Computational & Mathematical Organization
Theory, vol. 11, no. 2, pp. 97-118.
Lee, C. H., Yang, H.C., Chien, T.F. and Wen, W.S. (2011)
"A Novel Approach for Event Detection by Mining
Spatio-temporal Information on Microblogs",
Advances in Social Networks Analysis and Mining
(ASONAM), 2011 International Conference onIEEE, ,
pp. 254.
Lipparini, A. and Sobrero, M. (1994) "The glue and the
pieces: entrepreneurship and innovation in small-firm
networks", Journal of Business Venturing, vol. 9, no.
2, pp. 125-140.
Masurel, E. and Janszen, R. (1998) "The relationship
between SME cooperation and market concentration:
Evidence from small retailers in the Netherlands",
Journal of Small Business Management, vol. 36, pp.
68-73.
Michaelidou, N., Siamagka, N. T. and Christodoulides, G.
(2011) "Usage, barriers and measurement of social
media marketing: An exploratory investigation of
small and medium B2B brands", Industrial Marketing
Management.
Pak, A. and Paroubek, P. (2010) "Twitter as a corpus for
sentiment analysis and opinion mining", Proceedings
of LREC.
Papacharissi, Z. (2009) "The virtual geographies of social
networks: a comparative analysis of Facebook,
LinkedIn and ASmallWorld", New Media & Society,
vol. 11, no. 1-2, pp. 199-220.
Pitt, L., van der Merwe, R., Berthon, P., Salehi-Sangari, E.
and Caruana, A. (2006) "Global alliance networks: A
comparison of biotech SMEs in Sweden and
Australia", Industrial Marketing Management, vol. 35,
no. 5, pp. 600-610.
Shirzad, S. R., Bell D. (2013) "A Systematic Literature
Review of Flexible E-Procurement Marketplace", .
Skeels, M. M. and Grudin, J. (2009) "When social
networks cross boundaries: a case study of workplace
use of facebook and linkedin", Proceedings of the
ACM 2009 international conference on Supporting
group workACM, , pp. 95.
Thelwall, M., Buckley, K., Paltoglou, G., Cai, D. and
Kappas, A. (2010) "Sentiment strength detection in
short informal text", Journal of the American Society
for Information Science and Technology, vol. 61, no.
12, pp. 2544-2558.
Venkatesh, V. and Davis, F. D. (2000) "A theoretical
extension of the technology acceptance model: Four
longitudinal field studies", Management science, vol.
46, no. 2, pp. 186-204.
Walter, A., Ritter, T., Germunden, H. and Gemunden, H.
(1997) Relationships and networks in international
markets, Pergamon Press.
Yoo, J. and Hwang, J. (2008) "A framework for
discovering spatio-temporal cohesive networks",
Advances in Knowledge Discovery and Data Mining, ,
pp. 1056-1061.
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
570
