A Node-Merging based Approach for Generating iStar Models from User Stories

Abstract

User story is a widely adopted requirement notation in agile development. Generally, user stories are written by customers or users in natural language with limited form to describe user's needs for the software system from their perspectives. However, since user stories are generally presented in a flat list, the relations derived from the user stories are difficult to capture. It reduces the understanding of the system as a whole. One solution to this problem is to build goal-oriented models that provide explicit relations among user stories. But extracting concepts and relationships from a large number of discrete user stories often take a lot of time for the agile development team. This paper proposes an iStar model generating approach based on node-merging from user stories. The method first extracts the iStar nodes from the semi-structured user stories, then uses a BERT (Bidirectional Encoder Representations from Transformers) model to measure the similarity between the nodes, and then nodes to be merged are identified and the edges between the iStar nodes are connected. Experiments are designed to illustrate the effectiveness of the proposed approach.