In order to explore the effect of lexical semantic network construction on Chinese college students’ English reading comprehension, this paper selects 40 reading texts from 286 non-English major college students to construct a corpus. After preprocessing, 2316 valid words are preserved, and 4876 valid semantic relationship edges are formed. This paper constructs a lexical semantic network by means of natural language processing, semantic similarity calculation and graph model, and uses a reading test to test its influence on reading performance, word sense inference and discourse integration. The results show that the average clustering coefficient of the network is 0.312, the average path length is 5.27, and the proportion of the largest connected subgraph is 88.51%, which indicates that the lexical relations have obvious aggregation and connectivity. With the improvement of semantic network structure, the total score of reading increased from 66.3 to 82.1, the score of word meaning inference increased from 62.7 to 78.9, and the score of discourse integration increased from 64.1 to 81.4. The results show that lexical semantic network can effectively improve reading comprehension performance by enhancing semantic activation, inter-sentence cohesion and discourse construction, and provide a new method basis for college English reading teaching and intelligent reading support.
