Similarity Measures in Meaning Representation

Similarity measures in meaning representation

Maria Montefinese and David Vinson

UCL Institute for Multimodal Communication

Department of Experimental Psychology

(Ethical approval already obtained)

Semantic representation –knowledge of objects, word meanings, facts and people, without connection to any particular time or place– is a fundamental component of our mind. In literature, two main contrasting approaches to the study of semantic representation have been proposed. On the one hand, experiential cognition theories treat meaning as centrally based upon perceptual, motor and affective states arising from our direct sensory experience and actions (Martin, 2016). On the other hand, distributional theories treat meaning as a consequence of the statistical distribution of words across spoken and written language (see Andrews, Vigliocco & Vinson, 2009). From both approaches, different kinds of semantic similarity measures between concepts –i.e., the degree to which two concepts are close in their meaning– have been derived. These similarity measures are essential to our semantic representation. In fact, we widely use them every day to compare semantic entities, objects, actions, on the basis of their meaning. Although many research efforts have been made to understand humans’ semantic knowledge (see for example, Carlson et al., 2014), the nature of concept representations, that is, which similarity measure is used to represent meaning, still remains a hot topic of debate in the field.

The goal of this study is to investigate whether the similarity structure of abstract and concrete concepts based on neural and behavioural data reflects patterns of semantic similarity measures across concepts.To this aim, first, we will derive semantic similarity dimensions based on both experiential and distributional data. Second, healthy native English speakers will be enrolled in behavioural studies (i.e., lexical and semantic decision tasks) to investigate potential similarities in the representations based on behavioural and experiential/distributional similarity measures.Finally, healthy native English speakers will be enrolled in anfMRI study. During the fMRI session, participants will perform the same experimental tasks used in the behavioural studies –in separate short blocks– onvisuallypresented words denoting abstract and concrete concepts belonging to several semantic categories. Individual projects will include collaborating with us to develop semantic similarity dimensions and conducting behavioural or fMRI studies depending on student expertise and interest.

Suggested reading:

1)Martin, A. (2016). GRAPES—Grounding representations in action, perception, and emotion systems: How object properties and categories are represented in the human brain. Psychonomic Bulletin & Review, 23(4), 979-990.

2)Andrews, M., Vigliocco, G., & Vinson, D. (2009). Integrating experiential and distributional data to learn semantic representations. Psychological Review, 116(3), 463.

3)Carlson, T. A., Simmons, R. A., Kriegeskorte, N., & Slevc, L. R. (2014). The emergence of semantic meaning in the ventral temporal pathway. Journal of Cognitive Neuroscience, 26(1), 120-131.

Prerequisites:

Desirable:experimental psychology or semantic/linguistics background.

Required commitment:

Tasks include assisting in designing the behavioural/fMRIexperiments, recruiting participants, conducting the experiments and carrying out analysis (with support).

Contact:

Maria Montefinese:

David Vinson: