Topical text network construction based on seed word augmentation

A topical text network is helpful when analyzing a large corpus of documents, for it can give an intuitive insight of the topic distribution, including topic words and their connections. This paper proposes a topical text network construction method based on seed word augmentation. Firstly, we manually select some representative seed words for each topic, and then these words are augmented by some similarity metric. Secondly, by defining a threshold of similarity, the similar topic words are connected to construct a topical text network. We construct a topical network on aviation safety reports to analyze human factors using this method, and find the network a scale-free and small-world network with significant modularity.