Examining the robustness of web co‐link analysis
| Date | 25 September 2009 |
| DOI | https://doi.org/10.1108/14684520911001936 |
| Pages | 956-972 |
| Published date | 25 September 2009 |
| Author | Liwen Vaughan,Juan Tang,Jian Du |
| Subject Matter | Information & knowledge management,Library & information science |
Examining the robustness of web
co-link analysis
Liwen Vaughan
Faculty of Information and Media Studies, University of Western Ontario,
London, Canada, and
Juan Tang and Jian Du
Institute of Scientific and Technical Information of Shanghai, Shanghai, China
Abstract
Purpose – The purpose of this paper is to examine the robustness of web co-link analysis for
business intelligence.
Design/methodology/approach – The method is tested in two different Chinese industries, the
electronics/IT industry and the chemical industry. Web co-link data are collected in two different time
periods from a different search engine in each period. Multidimensional scaling (MDS) is used to map
the co-link data into business competition positions.
Findings – Web co-link analysis is fairly robust in that the mapping results reflect fairly well the
business competition landscape for both industries and in both time periods. The mapping results are
better when the data collection is restricted to Chinese language webpages only. The study also finds
that the Chinese webpages are very consumer-oriented, a phenomenon that is not seen in previous
studies of international companies.
Originality/value – This paper contributes to the understanding of the robustness and applicability
of the co-link analysis method. The method is useful for business intelligence and can also be applied
to the non-business environment. The paper also contributes to the understanding of a specific Chinese
web phenomenon.
Keywords Worldwide web, Intelligence, China
Paper type Research paper
Introduction
It has been well established that web hyperlinks contain useful information that can be
explored for various purposes. For example, Google (Page et al., 1998) an d other major
search engines (Sullivan, 2007) all use sophisticated algorithms to rank webpages
based on the ways that webpages link to each other. Patterns of webpage linking have
been used to identify web communities (Flake et al., 2002), to obtain scientific research
information (Thelwall, 2005) and business information (Reid, 2003; Thuraisingham,
2003), and to carry out counter-terrorism research (Mooney et al., 2004).
In the business world, the websites of business competitors tend not to link to each
other (Vaughan et al., 2006), but the website of a third party, such as a customer or a
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1468-4527.htm
This study was part of a larger project funded by the Initiative on the New Economy (INE)
Research Grants programme of the Social Sciences and Humanities Research Council of Canada
(SSHRC). Research assistants Karl Fast and Gloria Liu helped with data collection. Senior
chemical engineer Lei Xu of the Science and Technology Information Institute of Shanghai
Chemical Industry helped with the interpretation of the results of the chemical industry.
OIR
33,5
956
Refereed article received
20 October 2008
Approved for publication
30 January 2009
Online Information Review
Vol. 33 No. 5, 2009
pp. 956-972
qEmerald Group Publishing Limited
1468-4527
DOI 10.1108/14684520911001936
retailer, may link to a pair of business competitors, that is the websites of two business
competitors may be co-linked by a third party. In analysing the co-link patterns of
international telecommunications companies, Vaughan and You (2005a) found
business information in the form of a business competition map. The current study
applied the method developed in that earlier study to the Chinese environment to
determine if the method is applicable to an environment that has a different language,
culture and business practice. The testing was carried out in two different timeperiods,
Winter (February and March) 2007 and Fall (September) 2008. The results of the
Winter 2007 testing were reported in an earlier paper (Vaughan et al., 2008).
The overall approach of the study was to select a group of companies, locate
company websites and for each possible pair of companies use a search engine to find
all webpages that had hyperlinks that pointed to the pair of company websites (i.e. find
the number of co-links of these two companies), then construct a co-link matrix of all
the companies in the group and analyse the co-link matrix using a statistical method
called multidimensional scaling (MDS). The result of MDS is a map that attempts to
position all companies according to their similarities, with similar companies
positioned closer to each other. The similarity of a pair of companies is measured by
the number of co-links they have. The more co-links pointing to the two companies, the
more similar they are. This is based on the idea that the more two companies are
related, the more likely it is that they will be co-linked. For example, the website of a
computer retail store is likely to link to two computer companies that are their
suppliers, but it is unlikely that the website will have links to a computer company and
a food company. As similar or related companies are likely to be competitors (two
computer companies are competitors but a computer company and a food company are
not competitors), the MDS map will effectively place competing companies tog ether.
Thus the MDS map will in effective show the business competition landscape.
China’s two major industries, the electronics/information technology industry and
the chemical industry, were chosen for the study. These two Chinese industries are
major international players in their respective markets in that they are both major
exporters and importers. They are both a main focal point of their respective
international competitors. These two industries are also very different so they can be
used as contrasts to test the co-link analysis method. Each industry has thousands of
companies – too many to be included in the study, therefore it was decided to study the
top companies because reliable business information on these companies is more
readily available. Details of the company selection are reported in the Methodology
section below.
China was chosen for several reasons. First, there is a great interest in business
intelligence in China, as shown by the existence of a national organisation, the Society
of Competitive Intelligence of China (www.scic.org.cn), and several commercial portals
on competitive intelligence such as China’s Network of Competitive Intelligence
(www.chinaci.com).
Second, an examination of research literature showed a lack of research on business
intelligence methods in the Chinese environment. In China, there is plenty of discussion
of business intelligence but relatively little has been done to develop original research
methods for business intelligence. Although the methods used to produce business
intelligence reports are very rigorous, they tend to be traditional methods without the
use of web resources (China’s Network of Competitive Intelligence, 2007).
Robustness of
web co-link
analysis
957
To continue reading
Request your trial