Linguistics talk (2.3) 2PM by Prof. Hamad Al-Azary
Please join us for a linguistics talk by Prof. Hamad Al-Azary of Lawrence Tech on Feb. 3 (Friday) 2PM in 5057 Woodward Rm. 10302, a reception will follow.
You can also join us on Zoom (Meeting ID: 962 1641 6021, passcode: 051953)
Modelling Metaphorical Meaning: The Role of Semantic Richness
In a metaphor, such as language is a bridge, two distinct concepts are juxtaposed to create a novel understanding such that language connects people. Although people comprehend metaphors with ease, metaphor remains a significant challenge for artificial intelligence models. In this talk, I will focus on an established and psychologically plausible computational model of language comprehension known as the "predication algorithm". I will demonstrate that, although the predication algorithm outperforms alternative algorithms, it nonetheless has several shortcomings with metaphor processing. For example, the predication algorithm fails to distinguish between canonical metaphors (e.g., language is a bridge) and nonsensical reversals (e.g., a bridge is language). Moreover, the algorithm oftentimes misinterprets metaphors. Critically, I will argue that one reason the algorithm fails is because it does not include multiple semantic variables that influence metaphor comprehension, which I have identified in my research program. These semantic variables are associated with human linguistic and embodied experience and contribute to a word's semantic richness. Some words are semantically rich because they denote concepts that are concrete and easy to imagine (e.g., pen) whereas others are less-rich because they denote concepts that are abstract and difficult to imagine (e.g., idea). Moreover, some concrete words are particularly rich because they denote concepts that are easy to physically interact with, such as bicycle, whereas less-rich words denote things that are difficult or impossible to interact with, such as rainbow (i.e., Body-Object Interaction). In addition, words can be rich if they have many semantic neighbors, or less-rich if they have few semantic neighbors (i.e., Semantic Neighborhood Density). I will review a series of psycholinguistic and neuropsychological experiments I conducted that demonstrate semantic richness is in fact detrimental to metaphor processing. Finally, I will relate my experimental results to the predication algorithm and discuss ways the algorithm can be bolstered by factoring in semantic richness in its computations.
Hamad Al-Azary is an Assistant Professor of Psychology at Lawrence Technological University. As an experimental psychologist interested in language and cognition, he holds a PhD in psychology from The University of Western Ontario along with a BA and MSc from the University of Windsor and has received post-doctoral training at the University of Alberta. His main research is on metaphor processing, conceptual representations and abstraction and he uses multiple research methods, such as psycholinguistic experiments, computational modelling, neuropsychological case studies, discourse analysis, and cognitive neuroscience techniques (EEG; neuromodulation).