Term co-occurrence-based corpus network

This paper draws the term nodes and their links in corpus network based on term co-occurrence, and resultantly grasps its properties, using the mathematical graphs and the network-related physical measures. The topological processing of singular value decomposition (SVD), which compresses a large amount of co-occurrence information into a much smaller space, is similar to the common feature of feed-forward neural network systems where a large number of inputs are connected to a fairly small number of hidden layer nodes. Until now, the diagonal values by SVD have not been interpreted in full and the statistical step of hidden nodes depends only on the input. It is supposed that this study will provide the visual clues, which were described in the corpus, of the contraction-expansion homeostasis of the corpus network, and will help for actualizing the ambiguous axis in a space generated by SVD and characterizing the more efficient input-dependant steps of hidden nodes.