An artificial neural network that responds to written words like our brain's word form area

Interesting work from Dehaene and collaborators:  

Significance

Learning to read results in the formation of a specialized region in the human ventral visual cortex. This region, the visual word form area (VWFA), responds selectively to written words more than to other visual stimuli. However, how neural circuits at this site implement an invariant recognition of written words remains unknown. Here, we show how an artificial neural network initially designed for object recognition can be retrained to recognize words. Once literate, the network develops a sparse neuronal representation of words that replicates several known aspects of the cognitive neuroscience of reading and leads to precise predictions concerning how a small set of neurons implement the orthographic stage of reading acquisition using a compositional neural code.

Abstract

The visual word form area (VWFA) is a region of human inferotemporal cortex that emerges at a fixed location in the occipitotemporal cortex during reading acquisition and systematically responds to written words in literate individuals. According to the neuronal recycling hypothesis, this region arises through the repurposing, for letter recognition, of a subpart of the ventral visual pathway initially involved in face and object recognition. Furthermore, according to the biased connectivity hypothesis, its reproducible localization is due to preexisting connections from this subregion to areas involved in spoken-language processing. Here, we evaluate those hypotheses in an explicit computational model. We trained a deep convolutional neural network of the ventral visual pathway, first to categorize pictures and then to recognize written words invariantly for case, font, and size. We show that the model can account for many properties of the VWFA, particularly when a subset of units possesses a biased connectivity to word output units. The network develops a sparse, invariant representation of written words, based on a restricted set of reading-selective units. Their activation mimics several properties of the VWFA, and their lesioning causes a reading-specific deficit. The model predicts that, in literate brains, written words are encoded by a compositional neural code with neurons tuned either to individual letters and their ordinal position relative to word start or word ending or to pairs of letters (bigrams).