Join the Natural Language Processing Working Group for a talk from Dr. Trevor Cohen on Probing the Dimensions of Meaning in Medicine with Distributed Representations of Biomedical Language.
Distributed representations of language – referred to as semantic vectors, or text embeddings - have become the predominant mode of language representation in computational linguistics. As continuous vector representations, they fit naturally as components of neural network models, and their rise in popularity has accompanied the adoption of deep neural networks across problem domains.
Distributed representations have several advantages over discrete representational alternatives including natural measures of semantic relatedness, resilience to encoding errors, and convenient ways to bring information from outside corpora to bear on language processing tasks. This talk will highlight these and other practical advantages of distributed representations with illustrative examples drawn from recent and ongoing projects.
Application areas under discussion will range from detection of linguistic manifestations of changes in mental state and status to plain language summarization of the biomedical literature, with an emphasis on the unifying theme of the utility of distributed representations in situations in which the same ideas are expressed in different words.
Watch the Recording