Loglinear and multidimensional scaling models of digital library navigation

BUTTENFIELD, BARBARA P.; Reitsma, René

doi:10.1016/s1071-5819(02)90383-6

Cited by 11 publications

(15 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In many practical applications in market research, psychology, sociology, environmental research, genomics, and information retrieval for the Web and other document databases, data consist of similarity or dissimilarity measures on each pair of objects (Young 1987;Schutze and Silverstein 1997;Tibshirani et al 1999;Buttenfield and Reitsma 2002;Condon et al 2002;Courrieu 2002;Elvevag and Storms 2003;Priem, Love, and Shaffer 2002;Welchew et al 2002;Ren and Frymier 2003). Examples of such data include the co-purchase of items in a market, disagreements between votes made by pairs of politicians, the number of links between pairs of Web pages, the existence or intensity of social relationships between pairs of families, and the overlap of university applications by high school graduates.…”

Section: Introductionmentioning

confidence: 99%

Model-based Clustering with Dissimilarities: A Bayesian Approach

Oh¹,

Raftery²

2003

View full text Add to dashboard Cite

A Bayesian model-based clustering method is proposed for clustering objects on the basis of dissimilarites. This combines two basic ideas. The first is that the objects have latent positions in a Euclidean space, and that the observed dissimilarities are measurements of the Euclidean distances with error. The second idea is that the latent positions are generated from a mixture of multivariate normal distributions, each one corresponding to a cluster. We estimate the resulting model in a Bayesian way using Markov chain Monte Carlo. The method carries out multidimensional scaling and model-based clustering simultaneously, and yields good object configurations and good clustering results with reasonable measures of clustering uncertainties. In the examples we study, the clustering results based on low-dimensional configurations were almost as good as those based on high-dimensional ones. Thus, the method can be used as a tool for dimension reduction when clustering high-dimensional objects, which may be useful especially for visual inspection of clusters.We also propose a Bayesian criterion for choosing the dimension of the object configuration and the number of clusters simultaneously. This is easy to compute and works reasonably well in simulations and real examples.

show abstract

Section: Introductionmentioning

confidence: 99%

Model-based Clustering with Dissimilarities: A Bayesian Approach

Oh¹,

Raftery²

2003

View full text Add to dashboard Cite

show abstract

“…Unlike a priori approaches, a posteriori approaches make no assumptions about dimensionality and metric but instead try to resolve the geometry from secondary data with the help of ordination techniques such as factor analytic approaches (e.g., Doré & Ojasoo, 2001; Newby, 2001) or multidimensional scaling (Skupin, 2000, 2002; Buttenfield & Reitsma, 2002).…”

Section: Introduction: Spatialization Of Information Systemsmentioning

confidence: 99%

“…In one such a posteriori approach, Buttenfield and Reitsma (2002) propose to resolve geometry from the information latent in an information system's transaction log. They hypothesize that, under the assumption that system users navigate a space, their navigational record contains sufficient information to reconstruct that space.…”

Section: Introduction: Spatialization Of Information Systemsmentioning

confidence: 99%

Spatialization of Web sites using a weighted frequency model of navigation data

Reitsma

Thabane

MacLeod

2003

J. Am. Soc. Inf. Sci.

View full text Add to dashboard Cite

A common problem in the spatialization of information systems is the determination of geometry; i.e., dimensionality and metric. Such geometry is either chosen a priori or is inferred a posteriori from secondary data. Recent work emphasizes the use of geometric information latent in a system's navigational record. Resolving this information from its noisy background, however, requires an unambiguous criterion of selection. In this paper we use a previously published, statistical method for resolving a Web-based information system's geometry from navigational data. However, because of the method's (theoretical) sensitivity to data selection, a weighted frequency correction based on empirical probability distributions is applied. The effect of this correction on the Web-space geometry is investigated. Results indicate that the inferred geometry is robust; i.e., it does not significantly change under this probabilistic correction.

show abstract

“…With the advance of the World Wide Web and the availability of large information and document spaces [15], [21], [22] and digital libraries [6], [10], new applications in need of area partitioning have emerged; in particular the construction and delineation of information spaces [1], [13], [17], [18], [23], [26], [39], [40], [44], [47], [48].…”

mentioning

confidence: 99%

“…are located. Chen [15], Börner et al [11] and Dodge and Kitchin [22] discuss a variety of methods to accomplish this, ranging from traditional factor analytic [23], [41] and multidimensional scaling models [13], [47] to two-and three-dimensional tree maps [4], [5], [12], [25], [28], [46] and self-organizing maps [16], [17], [33] - [36], [48]. Several of these approaches not only address the reconstruction of the spaces and the placement of information items in them, but also attempt to partition the resultant space into distinct regions, mostly so that the regions' areas reflect the magnitude or extent of their contents; for instance, expressed as the number of items they contain.…”

mentioning

confidence: 99%

Weight-proportional Space Partitioning Using Adaptive Voronoi Diagrams

2007

View full text Add to dashboard Cite

Traditional application of Voronoi diagrams for space partitioning results in Voronoi regions, each with a specific area determined by the generators' relative locations and weights. Particularly in the area of information space (re)construction, however, there is a need for inverse solutions; i.e., finding weights that result in regions with predefined area ratios. In this paper, we formulate an adaptive Voronoi solution and propose a raster-based optimization method for finding the associated weight set. The solution consists of a combination of simple, fixed-point iteration with an optional spatial resolution refinement along the regions' boundaries using quadtree decomposition. We present the corresponding algorithm and its complexity analysis. The method is successfully tested on a series of ideal-typical cases and the interactions between the adaptive technique and boundary resolution refinement are explored and assessed.

show abstract

Loglinear and multidimensional scaling models of digital library navigation

Cited by 11 publications

References 19 publications

Model-based Clustering with Dissimilarities: A Bayesian Approach

Model-based Clustering with Dissimilarities: A Bayesian Approach

Spatialization of Web sites using a weighted frequency model of navigation data

Weight-proportional Space Partitioning Using Adaptive Voronoi Diagrams

Contact Info

Product

Resources

About