OntoL Product – Case Study #6: Project Info Query and Risk Analysis

Project Context

The client had previously spent 100,000 CNY on an AI solution (Dify) that delivered poor results. The author’s colleague had built the project‑management, contract‑management, and payment‑management systems, so they were familiar with part of the database schema and needed to infer the remaining business logic from the tables.

Data Understanding

Database documentation showed a mix of well‑commented tables and auto‑generated tables whose names ended with the suffix “编码”. Example table names included FORMTABLE_MAIN_105 and FORMTABLE_MAIN_104. The lack of documentation for many tables made the business analysis difficult.

Scope Definition

Because the project was small, the author limited the modeling effort to contract‑payment records rather than creating a company‑wide semantic model. The three functional goals were:

Query projects, related contracts, and payment fulfillment status.

View approval records with historical notes to remind downstream reviewers of important considerations.

Combine domain knowledge, regulations, and standards to identify risks, potentially referencing external data.

Ontology Generation

The author used the free large‑language model Qianwen (千问) to generate a complete ontology specification. The generated specification was then imported into the OntoL product.

Since the client did not provide real business data, a demo dataset was created and later refined.

Database Configuration

OntoL supports configurable data sources and custom SQL, so the author configured the Oracle database as the data source through the product’s UI.

Schema Exploration

After establishing the connection, the raw tables (e.g., FORMTABLE_MAIN_105, FORMTABLE_MAIN_104) became visible, highlighting the difficulty caused by missing documentation.

The schema visualisation displayed two structures:

Abox : entity instances.

Bbox : class definitions.

Model Enrichment

Necessary dynamic data generation functions and custom DSL functions were added to the ontology to support the required queries and risk‑identification logic.

Interactive Adjustment

With all configurations complete, the system allowed dialogue‑based adjustments of the ontology. The author plans further iterations to refine the model.