DBLPLink 2.0 -- An Entity Linker for the DBLP Scholarly Knowledge Graph

TL;DR

This paper presents DBLPLink 2.0, a zero-shot entity linker for the DBLP scholarly knowledge graph that now includes the new dblp:Stream entity type. It builds a full architecture around prompted LLMs for mention extraction, type-aware candidate retrieval via Elasticsearch, and KG neighborhood expansion, with per-triple LLM log-probability scoring to re-rank candidates without KG-embedding retraining. The authors implement a web demo and report comparative results across multiple LLMs, showing that a balance such as Qwen-2.5-3B yields strong performance and that incorporating neighborhood context improves guidance beyond text-only matching. Limitations stem from dataset availability for the current DBLP schema and the inability to directly compare with DBLPLink 1.0; future work targets collecting a dedicated dataset to enable deeper evaluation of the new Stream entities.

Abstract

In this work we present an entity linker for DBLP's 2025 version of RDF-based Knowledge Graph. Compared to the 2022 version, DBLP now considers publication venues as a new entity type called dblp:Stream. In the earlier version of DBLPLink, we trained KG-embeddings and re-rankers on a dataset to produce entity linkings. In contrast, in this work, we develop a zero-shot entity linker using LLMs using a novel method, where we re-rank candidate entities based on the log-probabilities of the "yes" token output at the penultimate layer of the LLM.

Figures (2)

• Figure 1: The DBLPLink 2.0 Web Interface
• Figure 2: The "Final Linked Results" Tab when expanded