Word Activation Forces Map Word Networks

Words associate with each other in a manner of intricate clusters 1-3. Yet the brain capably encodes the complex relations into workable networks 4-7 such that the onset of a word in the brain automatically and selectively activates its associates facilitating the language understanding and generation 8-10. One believes that the activation strength from one word to another forges and accounts for the latent structures of the word networks. This implies that mapping the word networks from brains to computers 11,12 , which is necessary for various purposes 1,2,13-15 , may be achieved through modeling the activation strengths. However, although a lot of investigations on word activation effects have been carried out 8-10,16-20 , modeling the activation strengths remains open. Consequently, huge labor is required to do the mappings 11,12. Here we show that our found word activation forces, statistically defined by a formula in the same form of the universal gravitation, capture essential information on the word networks, leading to a superior approach to the mappings. The approach compatibly encodes syntactical and semantic information into sparse coding directed networks, comprehensively highlights the features of individual words. We find that based on the directed networks sensible word clusters and hierarchies can be efficiently discovered. Our striking results strongly suggest that the word activation forces might reveal the