InfoQ article reduces financial analysis ontology to four elements and five build routes
An InfoQ Chinese AI & LLM column by Zhang Yang, a senior engineer at Ping An Life, argues that in financial data-questioning scenarios the key to making AI genuinely read business meaning and execute queries precisely is an ontology: business knowledge translated into machine-checkable structured language. The article says every ontology methodology comes down to four elements — concepts, attributes, relations and constraints (rules) — and that the industry's five construction routes (W3C RDF/OWL/SPARQL, top-level ontologies such as FIBO/BFO/DOLCE, architecture-driven TOGAF/ArchiMate, business-driven user stories/BPMN, and Palantir's Foundry/AIP) converge on the same core. It proposes three engineering tests — whether the AI can understand it, whether systems can execute it and whether the ontology can stay fresh — and states that its figures are illustrative and that no company operating or customer data is disclosed.
The piece offers enterprise practitioners a usability-first checklist for ontology work, aimed at financial query scenarios where the author says a pure LLM approach is most prone to silent errors in join and filter logic.

在金融问数场景下,要让 AI 真正"读懂"业务、精准执行查询,关键在于把业务知识翻译成机器可校验的结构化语言——本体。本文围绕两个问题展开:本体由哪些核心要素构成?行业有哪些建设路线、如何选型?核心结论:本体由"概念、属性、关系、约束(规则)"四要素构成,这也是所有方法论的内核;行业五条建设路线殊途同归,关键在贴合业务、能落地;判断标准是"AI 能不能听懂、系统能不能执行、本体能不能保鲜"三问。
一、本体的本质:把业务知识翻译成 AI 能读懂的语言
本体并非新概念,它在知识工程和语义网领域已有数十年积淀。哲学上它研究"存在",工程上它落地为一种可操作的建模方法。无论采用何种方法论,本体最终都落脚于四个核心要素。我们可以把它视为构建一个虚拟的"业务世界",这个世界由四类构件组成。理解这四类构件,就等于拿到了读懂一切本体模型的钥匙——无论它用的是 W3C 标准、顶层本体还是商业平台,内核都是这四个要素。
#Ping An Life#Palantir#Palantir Foundry#Palantir AIP#FIBO#BFO#DOLCE#TOGAF#ArchiMate