Scientific Web Intelligence

Introduction

Scientific Web Intelligence (SWI) is a research field that combines techniques from data mining, web intelligence and scientometrics to extract useful information from the links and text of academic-related web pages, using various clustering, visualization and counting techniques. Its origins lie in previous scientometric research into mining offline academic data sources such as journal citation databases, in contrast to Web Science, which focuses on engineering an effective Web (Berners-Lee et al., 2006). Typical scientometric objectives are either evaluative: assessing the impact of research; or relational: identifying patterns of communication within and between research fields. From scientometrics, SWI also inherits a need to validate its methods and results so that the methods can be justified to end-users and the causes of the results can be found and explained.

Background

The term ‘scientific’ in Scientific Web Intelligence has a dual meaning. The first meaning refers to the scope of the data: it must be academic-related. For example, the data may be extracted from university web sites, electronic journal sites, or just pages that mention or link to academic pages. The second meaning of ‘scientific’ alludes to the need for SWI research to use scientifically defensible techniques to obtain its results. This is particularly important when results are used for any kind of evaluation.

Scientific Web Intelligence is young enough that its basic techniques are not yet established (Thelwall, 2005c). The current emphasis is on methods rather than outputs and objectives. Methods are discussed in the next section. The ultimate objectives of typical developed SWI studies of the future can be predicted, however, from research fields that have used offline academic document databases for data mining purposes. These fields include bibliometrics, the study of academic documents, and scientometrics, the measurement of aspects of science, including through its documents (Borgman & Furner, 2002).

Evaluative scientometrics develops and applies quantitative techniques to assess aspects of the value of academic research or researchers. An example is the Journal Impact Factors (JIF) of the Institute for Scientific Information (ISI) that are reported in the ISI’s journal citation reports. JIFs are calculated for journals by counting citations to articles in the journal over a fixed period of time and dividing by the number of articles published in that time. Assuming that a citation to an article is an indicator of impact (because other published research has used the article in order to cite it), the JIF assesses the average impact of articles in the journal. By extension, ‘good’ journals should have a higher impact (Garfield, 1979), so JIFs could be used to rank or compare journals. In fact the above argument is highly simplistic. Scientometricians, whilst accepting the principle of citations as a useful impact proxy, will argue for more careful counting methods (e.g., not comparing citation counts between disciplines), and a much lower level of confidence in the results (e.g., taking them as indicative rather than definitive) (van Raan, 2000). Evaluative techniques are also commonly used for academic departments. For example, a government may use citation-based statistics in combination with peer review to conduct a comparative evaluation of all of the nation’s departments within a given discipline (van Raan, 2000). Scientific Web Intelligence may also be used in an evaluative role, but since its data source is only web pages, which are not the primary outputs of most scientific research, it is unlikely to ever be used to evaluate academics’ web publishing impact. Given the importance of the web in disseminating research (e.g., Lawrence, 2001), it is reasonable to measure web publishing, however.