Module structure detection of oracle characters with similar semantics

It is a highly productive approach to analyze human languages with complex networks, because the language network can effectively predict unknown features. One of the classical features in complex networks is module (or community) structure, which features dense intra-group connections and sparse inter-group connections. This work aims to identify oracle characters, the oldest characters in China, with similar semantic through module structure detection because about two-thirds of all oracle characters remain unknown. Firstly, two oracle character networks are constructed based on the context and shape of oracle rubbings and characters. Then, three module structure detection methods are employed to mine the modules from the oracle character networks. Through the experiments on the oracle character networks, three significant results are obtained. The first one is that oracle characters (or oracle variant characters) with similar semantics can be discovered by module structure, and the second one is that context and shape are of equal importance to the semantic formation of oracle characters. The last result is that modules have strong local topology of the network.