The invention discloses a
large model-based scientific research paper
structure analysis method and
system, and relates to the technical field of computers, and the method comprises the following steps: extracting reference links between paper segments by a
large model to construct a paper reference
dependency graph; detecting a closed-loop path on the
dependency graph, when the closed-loop path is detected, determining a starting point node and
copying to generate an expansion agent node,
cutting off a reference link of a
tail node pointing to the starting point node, changing to pointing to the expansion agent node, and converting the closed-loop path into a non-closed chain; replacing each closed-loop path to form a new-version paper reference
dependency graph, and modifying the paper
reference structure according to the new-version paper reference dependency graph to generate a modified scientific research paper; and inputting cue words into the
large model, and analyzing and outputting a result based on the modified scientific research paper. The problem of cyclic analysis caused by paper closed-loop reference during large model analysis is solved, the efficiency of the large model is improved, and the method is universal for various large models.