What is a neural model of a noun?

Today's published research on Predicting Human Brain Activity Associated with the Meanings of Nouns suggests that nouns activate both sensory-motor and planning networks in the brain. A noun activates the features of how it appears, tastes, and, some sort of plan. A plan of how to use the noun perhaps? The predictive model is far from perfect (70% where chance was rated at 50%), yet confirms that rather than understanding objects in a Platonic void, they are understood as things to be used, and much of their meaning is in how they are sensed and used.

While constructing a game that encourages the user to practice a foreign language, I'm wondering if a neural-inspired representation of nouns is useful. Object-oriented (object, property) and (object, method) data structures come to mind. Yet also behavior-oriented (event-driven, functional and aspect-oriented) representation comes to mind, if one considers that the actor and not the object is the determiner of how an object is used. Rather than The Sims, which encodes behavior of the agent in the object (the television tells the Sim how it will be used), the abilities and desires of the actor alter the meaning of the noun. Because of the need to unit-test and rapidly edit code in cross-cutting concerns behavior-oriented representation is already appealing. Computational linguistic representations of nouns and verbs, and associated architectures, strike me at first glance as daunting for constructing rapid prototypes. I suspect there is a lightweight middle-ground that could achieve some succinctness by modeling nouns according to agent usage, as humans do.

/science - permalink - comments