The semantic web gets a boost from functional MRIs (Lee @ Ars technica)
I'm recommending that you take a moment to read the above-linked interesting post over at Ars technica on an article recently published in Science. The post discusses the research in the article, which involved training a neural network to create a sense of meaning for a variety of nouns and verbs, and then linking the neural network's meaning for those words with fMRI brain scans of humans thinking about those words. The neural network was then challenged to predict fMRI patterns of brain activation associated with the words it had developed meanings for, but hadn't already linked with existing fMRI scans - it performed quite well, and continued to perform better than random chance would suggest as the prediction task became less linked to the original word set.
No comments:
Post a Comment