Extended similarity trees

Proximity data can be represented by an extended tree, which generalizes traditional trees by including marked segments that correspond to overlapping clusters. An extended tree is a graphical representation of the distinctive features model. A computer program (EXTREE) that constructs extended trees is described and applied to several sets of conceptual and perceptual proximity data.This research was supported in part by a National Science Foundation Pre-doctoral Fellowship to the first author.A magnetic tape containing both the EXTREE program described in the article and ADDTREE/P program for fitting additive trees can also be obtained from the above address. Requests for the program should be accompanied by a check for $25 made out to Teachers College, to cover the costs of the tape and postage.     

Additive similarity trees

Similarity data can be represented by additive trees. In this model, objects are represented by the external nodes of a tree, and the dissimilarity between objects is the length of the path joining them. The additive tree is less restrictive than the ultrametric tree, commonly known as the hierarchical clustering scheme. The two representations are characterized and compared. A computer program, ADDTREE, for the construction of additive trees is described and applied to several sets of data. A comparison of these results to the results of multidimensional scaling illustrates some empirical and theoretical advantages of tree representations over spatial representations of proximity data.We thank Nancy Henley and Vered Kraus for providing us with data, and Jan deLeeuw for calling our attention to relevant literature. The work of the first author was supported in part by the Psychology Unit of the Israel Defense Forces.     

Visualizing multiple word similarity measures

Although many recent advances have taken place in corpus-based tools, the techniques used to guide exploration and evaluation of these systems have advanced little. Typically, the plausibility of a semantic space is explored by sampling the nearest neighbors to a target word and evaluating the neighborhood on the basis of the modeler's intuition. Tools for visualization of these large-scale similarity spaces are nearly nonexistent. We present a new open-source tool to plot and visualize semantic spaces, thereby allowing researchers to rapidly explore patterns in visual data that describe the statistical relations between words. Words are visualized as nodes, and word similarities are shown as directed edges of varying strengths. The "Word-2-Word" visualization environment allows for easy manipulation of graph data to test word similarity measures on their own or in comparisons between multiple similarity metrics. The system contains a large library of statistical relationship models, along with an interface to teach them from various language sources. The modularity of the visualization environment allows for quick insertion of new similarity measures so as to compare new corpus-based metrics against the current state of the art. The software is available at www.indiana.edu/~semantic/word2word/ .     

Judging similarity among strings described by hierarchical trees

The paper compares the tree-theoretical model of similarity judgement (in which the similarity between two objects is a function of the distance between them in a conceptual tree) with an averaging model of similarity judgement that is drawn jointly from information integration theory and from current research indicating the prevalence of anchoring and adjustment mechanisms in judgement. Results of an experiment are presented that suggest that even when subjects organize conceptual material as a hierarchical tree, judgments of similarity among the objects are better accounted for by an averaging mechanism than by distances in the tree. These data are discussed in terms of the differences between the representation in which knowledge is encoded and the processes that operate on the represented information.     

Relations between premise similarity and inductive strength

According to the diversity principle, diverse evidence is strong evidence. There has been considerable evidence that people respect this principle in inductive reasoning. However, exceptions may be particularly informative. Medin, Coley, Storms, and Hayes (2003) introduced a relevance theory of inductive reasoning and used this theory to predict exceptions, including the nondiversity-by-propertyreinforcement effect. A new experiment in which this phenomenon was investigated is reported here. Subjects made inductive strength judgments and similarity judgments for stimuli from Medin et al. (2003). The inductive strength judgments showed the same pattern as that in Medin et al. (2003); however, the similarity judgments suggested that the pattern should be interpreted as a diversity effect, rather than as a nondiversity effect. It is concluded that the evidence regarding the predicted nondiversity-byproperty-reinforcement effect does not give distinctive support for relevance theory, although this theory does address other results.     

CoTO: A novel approach for fuzzy aggregation of semantic similarity measures

Semantic similarity measurement aims to determine the likeness between two text expressions that use different lexicographies for representing the same real object or idea. There are a lot of semantic similarity measures for addressing this problem. However, the best results have been achieved when aggregating a number of simple similarity measures. This means that after the various similarity values have been calculated, the overall similarity for a pair of text expressions is computed using an aggregation function of these individual semantic similarity values. This aggregation is often computed by means of statistical functions. In this work, we present CoTO (Consensus or Trade-Off) a solution based on fuzzy logic that is able to outperform these traditional approaches.     

Classification behavior and measures of intelligence: dimensional identity versus overall similarity

Individuals tend to adopt either analytic or holistic modes of categorizing objects. In two studies, we examined the relation between these categorization tendencies and cognitive abilities as measured by standard psychometric instruments. The participants in both studies were pretested with a restricted classification task in which it was possible for them to classify simple stimuli by dimensional identity or overall similarity. Those making a large number of either type of categorization were then tested with subtests of the WAIS-R and with the Raven's progressive matrices. Across both studies, the analytic individuals (many dimensional identity classifications) scored higher than the holistic individuals (many overall similarity classifications) on some but not all of the subtests. The results are consistent with the idea that holistic modes of categorization may be more "primitive" than analytic modes. The findings are discussed in terms of the association between categorization mode and either general or specific cognitive abilities.     

Content-specific phenomenological similarity between episodic memory and simulation

ABSTRACT

Numerous studies have indicated that remembering specific past experiences (i.e., episodic memory) and imagining specific novel future experiences (i.e., episodic simulation) are supported by common mental processes. An open question, however, is whether and to what extent the content of specific past episodes is sampled when simulating a specific future episode. The current study aimed to answer this question. Participants recalled past episodes each comprising two episodic details, a personally familiar location and person. Participants also simulated novel future episodes using recombined pairs of person and location details taken from different recalled episodes. Participants rated the vividness of each location and person in their memory and simulation. We conducted a multi-level analysis where the vividness rating during memory was used to predict the vividness rating during simulation at the level of individual shared details (i.e., location or person). The vividness of the memorial detail co-varied with the vividness of the simulated detail; this relationship persisted even after accounting for the underlying familiarity of the details. These findings strongly suggest that simulations of specific future experiences are based upon the contents of specific prior episodes.     

On the relation between similarity and ratio estimates

Summary Data obtained in four sets of experiments involving pitch, heaviness, greyness, and circular area were re-analyzed. It was found (1) that similarity estimates S _ijare a power function of stimulus ratios S _ijwith the exponent n _s,(2) that ratio estimates q _ijare also a power function of stimulus ratios S _ijwith the exponent n _q(i.e., Stevens' power law), (3) that the exponent n of similarity estimates as power function of ratio estimates is equal to the ratio n _s/n _q,and (4) that, inversely, the exponent m of ratio estimates as power function of similarities is equal to the ratio n _q/n _s.The investigation was supported by a grant from the Swedish Council for Social Science Research.