With the rapid development of information technology, the demand for retrieving massive amounts of data has been growing. Traditional retrieval algorithms face challenges in efficiency and accuracy when dealing with large-scale data. This paper proposes an intelligent retrieval algorithm that combines graph embedding and dynamic weight indexing to improve retrieval efficiency and accuracy. The algorithm captures the complex semantic relationships between documents and terms using graph embedding technology and optimizes the index structure with dynamic weight skip lists, combined with the BM25 algorithm for relevance scoring. Experimental results demonstrate that the proposed algorithm achieves significant performance improvements on large-scale datasets.