Automatic Detection of Causality in Requirement Artifacts: the CiRA Approach

TL;DR

This paper presents CiRA, a novel tool-supported approach for detecting causality in natural language requirements, achieving high accuracy and outperforming existing methods, thereby advancing automated reasoning and testing in requirements engineering.

Contribution

The paper provides the first comprehensive analysis of causality in requirements and introduces CiRA, a tool that effectively detects causal relations with high precision and recall.

Findings

Causality is present in about one-third of requirement sentences.

CiRA achieves a macro-F1 score of 82%.

It outperforms related approaches by over 11% in macro-Recall and macro-Precision.

Abstract

System behavior is often expressed by causal relations in requirements (e.g., If event 1, then event 2). Automatically extracting this embedded causal knowledge supports not only reasoning about requirements dependencies, but also various automated engineering tasks such as seamless derivation of test cases. However, causality extraction from natural language is still an open research challenge as existing approaches fail to extract causality with reasonable performance. We understand causality extraction from requirements as a two-step problem: First, we need to detect if requirements have causal properties or not. Second, we need to understand and extract their causal relations. At present, though, we lack knowledge about the form and complexity of causality in requirements, which is necessary to develop a suitable approach addressing these two problems. We conduct an exploratory case study with 14,983 sentences from 53 requirements documents originating from 18 different domains and shed light on the form and complexity of causality in requirements. Based on our findings, we develop a tool-supported approach for causality detection (CiRA). This constitutes a first step towards causality extraction from NL requirements. We report on a case study and the resulting tool-supported approach for causality detection in requirements. Our case study corroborates, among other things, that causality is, in fact, a widely used linguistic pattern to describe system behavior, as about a third of the analyzed sentences are causal. We further demonstrate that our tool CiRA achieves a macro-F1 score of 82 % on real word data and that it outperforms related approaches with an average gain of 11.06 % in macro-Recall and 11.43 % in macro-Precision. Finally, we disclose our open data sets as well as our tool to foster the discourse on the automatic detection of causality in the RE community.