Use case: Automated Analysis of Technical Specifications

Field

Railway

Type of company

Large Enterprise

The problem to be solved

The client needed to analyze technical specifications spanning thousands of pages. In this specific case, analyzing over 2,000 pages of technical specifications required four full-time engineers for at least five days. The technical and complex nature of the language, along with the poor structuring of the information, made the process inefficient. From a technical standpoint, there were no tools available to support the automatic extraction, evaluation, and structuring of requirements, leaving engineers to manually distinguish between truly relevant content and accessory information. The documents, mostly in PDF format, contained complex technical text, tables, and numerous regulatory references, with high variability in both lexical and structural terms, making it difficult to apply recurring or automatable analysis patterns.

The solution

KYD® was configured to perform automatic pre-processing of the specifications, aimed at identifying and extracting requirements and technical parameters directly from the text, significantly reducing the need for manual intervention. To support this process, a technical knowledge base was introduced, including units of measurement, physical properties, functions, and regulatory references. This component not only improved the system’s ability to interpret requirements, but also enabled a more objective assessment of their completeness and clarity. A key feature of the solution was the automatic calculation of an ambiguity score, designed to highlight potentially unclear or ambiguous requirements from the earliest stages of analysis. This allowed the technical team to immediately focus on the most critical parts of the document, making the entire process more efficient, targeted, and repeatable.

The measurable results

• Over 6,200 requirements extracted from more than 2,000 pages
• Identification of over 543 regulatory references
• Compilation of a list of 50+ regulatory acronyms previously unknown to the client
• 84% match with manual tagging by engineers, and over 200 additional requirements identified by the system
• Over 90% reduction in the time needed for requirement data entry in the client’s management software