Human Similarity Judgments Research Articles

Effects of Mug Book Size and Computerized Pruning on the Usefulness of Dynamic Mug Book Procedures

The usefulness of dynamic mug books with larger sets of mug shots was explored in experiment 1. Participants viewed a videotaped crime and attempted an identification from either a dynamic mug book (computerized video clips could be chosen) or a static mug book. The perpetrator appeared in one of three positions - 70, 140, or 210. When the perpetrator and the 69 preceding foils appeared in positions 1-70, correct identifications and false positives were higher than for later positions. Dynamic information usage decreased so dramatically for the later positions that it was argued that dynamic mug books should be used only when the number of pictures could be pruned. Experiment 2 explored the usefulness of pruning dynamic mug books. Identifications were attempted in mug books that contained either 69 pictures pruned by a computer facial-recognition algorithm for their similarity to the perpetrator, or a random selection of 69 pictures. As in past research, the dynamic mug book reduced false positives; however, in contrast to past research, similarity of mug book pictures as defined by the algorithm did not affect mug book performance. Experiment 3 tested the relationship between the computer facial-recognition algorithm similarity ratings used in experiment 2 with human similarity judgments. No relationship was found between human and computer ratings of similarity.

Semantic Similarity in a Taxonomy: An Information-Based Measure and its Application to Problems of Ambiguity in Natural Language

This article presents a measure of semantic similarity in an IS-A taxonomy based on the notion of shared information content. Experimental evaluation against a benchmark set of human similarity judgments demonstrates that the measure performs better than the traditional edge-counting approach. The article presents algorithms that take advantage of taxonomic similarity in resolving syntactic and semantic ambiguity, along with experimental results demonstrating their effectiveness.

Predicting face similarity judgements with a computational model of face space

Current models of face representation involve the notion of a high-dimensional face space. Computational models of face space based on principal components analysis (PCA) have been successfully used to predict human judgements of face sex or race. In this work the capability of PCA-based face spaces to predict human judgements of face similarity is examined. Three different paradigms were used. In Experiment 1 subjects learned face-name associations for 18 faces and identified these faces on tachistoscopic presentation. The number of confusions was used as a measure of face similarity. In Experiment 2 the same subjects subjectively rated the similarity of all 153 possible face pairs. In Experiment 3 reaction time to identify a face in an odd-man-out task was measured as an index of face similarity. These empirical measures were correlated with distance of the faces in PCA-based spaces of different dimensionalities. For Experiments 1 and 2 these correlations were highest for one-or two-dimensional face spaces ( r=−0.27 vs. −0.28). For Experiment 3 the correlation was highest for a space consisting of 13 dimensions ( r=−0.51). Thus PCA-based spaces seem capable to predict human similarity judgements to some extent. Possible reasons for the differences in predictability between paradigms are discussed.

Determining the similarity of deformable shapes.

Determining the similarity of two shapes is a significant task in both machine and human vision systems that must recognize or classify objects. The exact properties of human shape similarity judgements are not well understood yet, and this task is particularly difficult in domains where the shapes are not related by rigid transformation. In this paper we identify a number of possibly desirable properties of a shape similarity method, and determine the extent to which these properties can be captured by approaches that compare local properties of the contours of the shapes, through elastic matching. Special attention is devoted to objects that possess articulations, i.e. articulated parts. Elastic matching evaluates the similarity of two shapes as the sum of local deformations needed to change one shape into another. We show that similarities of part structure can be captured by such an approach, without the explicit computation of part structure. This may be of importance, since although parts appear to play a significant role in visual recognition, it is difficult to stably determine part structure. We also show novel results about how one can evaluate smooth and polyhedral shapes with the same method. Finally, we describe shape similarity effects that cannot be handled by current approaches.

Learning-based similarity measurement for fuzzy sets

The work described in this paper proposes a method for the measurement of similarity, viewed from the decision maker's perspective. At first, an algorithm is presented that generalizes a discrete fuzzy set F, representing a model, given another discrete fuzzy set G representing new evidence. The algorithm proceeds by expressing the fuzzy sets as possibility distributions, and then by extending the focal elements of F with elements from the focal elements of G, constructs a generalized fuzzy set. If the initial fuzzy model is repetitively updated with the same evidence, a convergence state will be reached, the number of repetitions depending on the relative position of the two sets. This number is considered to give an indication of the conceptual proximity of the two entities, and therefore of the two fuzzy sets, and can be used as a similarity index. This index is further complemented with a second measure, equal to the reduction of the probability assigned to the element with full membership in F, before learning and at the convergence state. This proposal for similarity seems to be in better agreement with experimental findings in human similarity judgment than approaches based on pointwise distance metrics. Q 1998 John Wiley & Sons, Inc.