Library knowledge discovery method and system based on semantic enhancement and large model cooperation

Abstract

This invention provides a library knowledge discovery method and system based on semantic enhancement and large-scale model collaboration. The method includes: extracting entities from each scientific document and mapping them to unified subject terms; constructing a continuous-temporal bipartite graph, where the node set contains document nodes and subject term nodes, and the edges in the edge set represent interaction events between document nodes and subject term nodes; enumerating all unordered subject term pairs for the subject term set of scientific documents; when a new target scientific document is added to the collection, generating a semantic message and updating the display memory state of nodes related to the target scientific document based on the semantic message; for nodes in each unordered subject term pair, aggregating the information of the current node in the spatiotemporal neighborhood to obtain temporal embeddings; predicting the probability of forming interaction events based on the temporal embeddings of unordered subject term pairs, and selecting a knowledge candidate set based on the probability; inputting the candidate associations, confidence scores, and corresponding contexts from the candidate set into a large language model to generate an intelligence analysis report.

Images