On the task of predicting the range of possible next words in a sentence, many networks (e.g. Elman, 1990) that have been proposed are capable of displaying a certain degree of systematicity, but fail in recognizing grammatically correct but semantically anomalous sentences. Based on an expansion of Hadley's model (Hadley et al, 2001), I present a competitive network, which employs two
... [Show full abstract] sub-networks that discern coarse- grained and fine-grained categories respectively, by being trained via different parameter settings. Hence, one of the sub-networks will have a greater capacity for recognizing the syntactic structure of the preceding words, while the other will have a greater capacity for recognizing the semantic structure. This corresponds to the recent suggestion about specialization of the two hemispheres in the human brain (Beeman, 1998). Also, a mechanism to switch attention between the predictions from the two sub-networks is employed in order to make the global network more closely approximate human behavior. The results show that the network is able to deal with grammatically correct but semantically anomalous sentences.