Document Identifier

Document identifiers (DocIDs) are crucial for efficient information retrieval, and recent research focuses on generating them directly from queries using generative models, bypassing traditional methods. This approach leverages autoregressive models and explores various identifier strategies, including numerical sequences, set-based identifiers from lexical tokens, and abstractive keyphrases generated by large language models. These advancements aim to improve retrieval speed and accuracy, leading to more efficient and effective search engines and knowledge-intensive applications.