The Synergy of Large Language Models and Knowledge Graphs: Current Status and Future Directions

Before large models, knowledge graphs already had many applications such as knowledge‑based question answering. When combined with search engines (e.g., knowledge cards), they provide direct factual answers and objective knowledge displays, and they remain more adept in this area. This mode will not be greatly impacted by the arrival of large models because knowledge acquisition scenarios are inherently broad.

Large models enhance human‑machine interaction by lowering the interaction threshold, making AI’s understanding of natural‑language input more thorough, detailed, and accurate. For casual user questions, large models are more flexible, and their answers become more natural.

Thus, large models and knowledge graphs are not a complete replacement relationship; they can be combined to solve problems previously unsolvable. Earlier knowledge‑graph‑based conversational systems involved chit‑chat, factual Q&A, and task‑oriented Q&A (e.g., booking flights or hotels), making the systems complex and heavy. After integrating large models, chit‑chat and factual Q&A can be better fused, resulting in more natural, end‑to‑end dialogue systems.

When to use knowledge graphs, large models will naturally find the optimal approach; task‑oriented Q&A can also be organically combined with other APIs, improving customer‑service and dialogue‑system architectures.

Moreover, knowledge graphs retain irreplaceable roles in highly structured domains such as risk control and financial market analysis, ensuring their long‑term existence.

Knowledge graphs have long faced a serious problem: high construction complexity and cost, lacking a unified building capability because many knowledge expressions are diverse and non‑standard in text or multimodal forms.

The industry is actively exploring how to leverage the powerful semantic understanding of large models to enhance standardized knowledge construction; progress has been made, yet many challenges remain.

Knowledge graphs need to lower costs; if costs become low enough, structured decisions will be more rigorous than those made by large models. Therefore, we must both accommodate large models and insist on the long‑term presence of knowledge graphs.

Consequently, these two paradigms should be integrated in an inclusive manner.

To deeply discuss the current status and future development of knowledge graphs and large models, DataFunSummit2024: Knowledge Graph Online Summit will be held online on March 23, 2024, 9:00‑17:00, welcoming practitioners to participate and exchange.