In this paper, a semantic network-based topic word extraction model for Chinese language features is constructed by combining the theoretical knowledge of the conceptual semantic network. Chinese language texts and videos on social networks are selected to build a lexicon and a knowledge base, and the forward maximum matching method is used for lexical segmentation. After calculating the weights of all the candidate subject words using the vector space model, the words are sorted from largest to smallest in terms of weight, and then the top N are filtered out as the final subject words to analyse their influence in cross-cultural communication. It can be seen that among the lexical features of the linguistic landscape, the word “shop” has the largest weight value of 0.2326, and among the lexical features of the linguistic landscape, “gold shop” and “medicine shop” are the most common. The most common word features in the linguistic landscape are “gold bank”, “medicine bank”, etc. The mean value of harmonic Chinese network buzzwords is 3.0078, which is the most widely spread. The number of likes and shares of video content without a series is generally high. In the event of Chinese language fever, it is important to investigate the means of international dissemination of Chinese language features based on semantic networks.