Research on Structured Extraction Method for Function Points Based on Event Extraction

Abstract

Software size is a significant input for software cost estimation, and the implementation of software size estimation dramatically affects the results and efficiency of cost estimation. Traditionally, the software size estimation is implemented by strictly trained experts and is more labor-intensive for large software projects, which is relatively expensive and inefficient. Function Point Analysis is a widely used method for software size estimation, supported by several international standards. We propose a structured and automated function point extraction method based on event extraction in natural language processing to address the problem of complex and inefficient manual recognition for function point recognition. This approach has been validated in 10 industrial cases. Experimental results show that our method can identify more than 70% of the function points, which significantly improves the efficiency of Function Point Analysis implementation. This paper could be a guide on the application of artificial intelligence techniques to software cost estimation.