InfoMetrics

Artificial Intelligence

Automated Function Point Analysis / Natural language processing

Designed and implemented software to establish Function Point metrics automatically from design documents describing business systems. The system reads documents written in natural language (Dutch).

This is essentially a text mining application, which processes text, then through the use of targeted statistics, accurately establishes logical objects and the transactions defined on these objects. The system very efficiently represents the knowledge gained from the texts, decides on subdomains, and uses analytics to decide which parts of a document are relevant to the domain, all without human assistance.

Significant reductions in time and effort and increases in consistency amongst human FP analysts using this software have been realized. Function Point analysis is especially useful to assess completeness of designs and to help establish plans and budgets for realization. The software system discussed is designed to be applied in early stages of the development cycle, when it is most useful. The system reduces the time needed for a full FP-analysis, including reporting, by a factor of three, compared to human performance.