How a leading car manufacturer
automated its requirement analysis process

Download PDF

Challenges

• The company was coping with a growing number of requirements (over 150,000) that were described in multiple documents and databases
• Thousands of unstandardized specification sheets existed, each of them containing on average 500 requirements partially hidden in complex tables
• There was no information about whether a requirement relates to a specific project or to multiple projects, nor an easy way to compare requirements across projects
• Since all automation efforts had failed, the requirement engineers had to copy-paste requirements and compare them manually

Solution

Leveraging Cortical.io patented approach to natural language understanding, the company developed a new requirement analysis tool that

• automates the grouping of technical requirements by comparing their semantic similarity across specifications
• outputs the requirements analysis in a table that lists all requirements, showing in which projects they occur, whether they are project-specific or not, and the value of similar properties like weight, material, color etc. across projects
• is accessible via a simple web interface

The Cortical.io Difference

• Because it does not rely on keyword analysis, the solution is able to match different formulations of the same requirement, even when the text is short or hidden in a table
• The solution was easily trained with reference data from the car manufacturer’ language domain (specification sheets, indices of abbreviations, company-specific standards, industry standards, automotive engineering literature, etc.)