A comparison of two techniques for bibliometric mapping: Multidimensional scaling and VOS (1003.2551v1)

Abstract: VOS is a new mapping technique that can serve as an alternative to the well-known technique of multidimensional scaling. We present an extensive comparison between the use of multidimensional scaling and the use of VOS for constructing bibliometric maps. In our theoretical analysis, we show the mathematical relation between the two techniques. In our experimental analysis, we use the techniques for constructing maps of authors, journals, and keywords. Two commonly used approaches to bibliometric mapping, both based on multidimensional scaling, turn out to produce maps that suffer from artifacts. Maps constructed using VOS turn out not to have this problem. We conclude that in general maps constructed using VOS provide a more satisfactory representation of a data set than maps constructed using well-known multidimensional scaling approaches.

• The paper establishes that VOS offers a more authentic mapping of bibliometric data than MDS by addressing distortions caused by zero-value similarities.
• The study rigorously evaluates three distinct data sets—information science authors, social science journals, and operations research keywords—to assess methodological impacts.
• The findings underscore the potential of VOS to minimize artificial spatial artifacts common in MDS, thereby enhancing the accuracy of scholarly visualizations.

A Comparative Analysis of Bibliometric Mapping Techniques: Multidimensional Scaling vs. VOS

The paper by Nees Jan van Eck, Ludo Waltman, Rommert Dekker, and Jan van den Berg offers an in-depth examination of two prominent techniques for constructing bibliometric maps: Multidimensional Scaling (MDS) and a novel method termed VOS (Visualization of Similarities). This paper provides both theoretical insights and empirical evaluations, elucidating the strengths and limitations of these methodologies in the visualization of bibliometric data.

MDS and VOS: Theoretical Underpinnings

MDS, a widely utilized technique in bibliometrics, aims to represent data in a low-dimensional space, such that the distance between any two items reflects their similarity. Although predominant in bibliometric mapping, MDS can exhibit certain limitations, particularly when dealing with similarity data characterized by a high incidence of zero values, leading to distortions in map representations.

Conversely, VOS is proposed as an alternative that leverages a weighted approach to the construction of bibliometric maps. The technique addresses several shortcomings of MDS by employing similarities measured on a ratio scale and minimizing a weighted sum of squared distances, which is designed to better preserve the relational structure of data.

Experimental Evaluation

The authors present a rigorous experimental comparison using three distinct bibliometric data sets:

1. Authors in Information Science: This data set elucidates connections within the information science field.
2. Journals in the Social Sciences: Encompassing a broader disciplinary range, this set highlights co-citation patterns on a journal level.
3. Keywords in Operations Research: Focusing on keyword co-occurrences, this data set is pertinent for mapping thematic areas within operations research.

Each data set was subjected to three mapping approaches: MDS-AS (using association strength similarity measure), MDS-COS (using cosine similarity), and VOS (using association strength).

Key Findings

The paper reveals that maps constructed via VOS provide a more authentic representation of data sets than those generated by MDS techniques. The MDS approaches were observed to frequently exhibit artifacts, such as a propensity to organize items in a circular manner or centralize prominent figures, regardless of their actual relational ties. Such tendencies can obscure the true structure of bibliometric data.

In contrast, the VOS approach demonstrated a robust ability to differentiate subfields without imposing artificial spatial structures. For instance, in the information science author map, VOS clearly delineated the informetrics and information seeking/retrieval subfields, a distinction less apparent in MDS maps.

Implications and Future Directions

The findings underscore VOS as a viable and, in many ways, superior alternative to MDS for bibliometric mapping. This has significant implications for researchers seeking to visualize scholarly landscapes, as VOS offers a more accurate tool that reduces common MDS pitfalls.

For future developments, it would be beneficial to explore VOS's integration with alternative similarity measures and its application to diverse bibliometric scenarios. As bibliometrics continues to evolve with advancing technologies and expanding data volumes, tools like VOS, with demonstrated precision and versatility, will be instrumental in mapping intellectual domains.

In conclusion, this paper lays a firm foundation for the transition towards more reliable bibliometric mapping techniques, proposing VOS as a noteworthy enhancement over traditional MDS approaches. This work invites further exploration and application in various research contexts, leveraging its robust framework to advance the accuracy and interpretability of scholarly visualizations.