ESTaid, which operates the search portal Zum (zum), said on the 4th that it has filed a patent application for a "knowledge-graph-based question-and-answer device and method."
The patent combines the generative capabilities of a large language model (LLM) with the structural reasoning of a knowledge graph to deliver accurate, evidence-based answers to complex questions. It is regarded as a core technology of the next-generation AI question-and-answer engine ESTaid is pursuing.
Conventional AI search has largely focused on document summarization, often causing the "hallucination" problem of generating information that differs from facts. To overcome this limitation, ESTaid designed the AI to analyze relationships among people, objects, and events in a query on its own and to explore information consolidation through a knowledge graph. As a result, it can present logical answers based on factual relationships beyond simple word similarity predictions.
When a user enters a query, the system extracts key concepts and actions to form a network of relationships and generates a knowledge graph based on it. The LLM then composes an answer from that graph, so the output contains highly reliable, data-driven information.
The technology shows superior performance over existing methods in areas such as ▲ improved accuracy for complex queries ▲ minimized LLM hallucinations ▲ secured scalability through continuous updates to graph data ▲ enhanced reliability by visualizing the basis for answers. It also introduces a "hybrid search" approach that combines semantic search and keyword search to more precisely reflect a user's search intent.
Based on this patented technology, ESTaid plans to advance its AI services overall, including real-time issue extraction, recommended query generation, and improved search accuracy. It also aims to strengthen its AI search competitiveness by boosting usability across various areas such as "AI issue analysis" and "customized information exploration services."
Kim Nam-hyeon, chief executive of ESTaid, said, "This technology is an important turning point for advancing from the stage where AI merely generates answers to 'intelligent AI search' that understands facts and intent and reasons logically," adding, "We will continue technological innovation so AI can deliver a more accurate and trustworthy information experience to users."