Modeling the neurophysiology of language
My work on modeling lexical recognition (see,
Moscoso
del Prado, 2008) has led to a collaboration with
researchers in the field of vision investigating how the model can
predict the eye movement and reading performance patterns both in
healthy adults and in patients with ocular macular degenarations (i.e.,
scotomata). In recent work
(see the
preliminary
version, and a more recent
poster)
we have developed a statistical model
capable of predicting the patterns of eye fixations (locations and
times) in reading for both patients and healthy subjects. Further work
in this direction (now in preparation), has produced a complete model
of reading that extends beyond the single word level into the plain
text level. This is remarkable, as it is the sole existing model of
human reading capable of integrating information from the level of a
visual image up to semantic processing, that is, text is presented to
the model in the most realistic version, a plain unlabeled bitmap, and
the model must recover all information all the way from features of
letters up to full sentences.