Semantic feature norms: a cross-method and cross-language comparison

Abstract

The ability to assign meaning to perceptual stimuli forms the basis of human behavior and the ability to use language. The meanings of things have primarily been probed using behavioral production norms and corpus-derived statistical methods. However, it is not known to what extent the collection method and the language being probed influence the resulting semantic feature vectors. In this study, we compare behavioral with corpus-based norms, across Finnish and English, using an all-to-all approach. To complete the set of norms required for this study, we present a new set of Finnish behavioral production norms, containing both abstract and concrete concepts. We found that all the norms provide largely similar information about the relationships of concrete objects and allow item-level mapping across norms sets. This validates the use of the corpus-derived norms which are easier to obtain than behavioral norms, which are labor-intensive to collect, for studies that do not depend on subtle differences in meaning between close semantic neighbors.

Keywords: Behavioral norms; Semantic features; Text corpora.

Fig. 1

Organization of the abstract target words. Visualization is based on a self-organizing map trained on the feature vectors of the abstract targets in Aalto norms. The self-organizing map is further divided using k-means clustering. The best clustering is selected using Davies–Bouldin index. A toroid map of 60 units was used. In the toroid shape, the units at the opposite edges of the sheet visualized are neighbors

Fig. 2

Dissimilarity matrices (cosine distance) of the 98 stimuli shared across the five data sets. The pairwise Spearman rank correlations are indicated in the figure

Fig. 3

Confusion matrix of the zero-shot learning models. The sum of misclassifications over all the 20 zero-shot learning models is shown. Zero is shown as white to ease interpretation