[Paper Review] On the Applicability of Explainable Artificial Intelligence for Software Requirement Analysis

Behnaz Jamasb, Reza Akbari+1 more|arXiv (Cornell University)|Feb 10, 2023

Software Engineering Research4 citations

TL;DR

This paper proposes an explainable AI (XAI) approach for software requirement classification using the LIME algorithm to enhance transparency in machine learning predictions. Evaluated on the PROMISE dataset, the method improves analyst understanding of classification decisions and aids in feature analysis, demonstrating XAI's value in requirement engineering with measurable benefits in interpretability and feature relevance.

ABSTRACT

The applications of Artificial Intelligence (AI) methods especially machine learning techniques have increased in recent years. Classification algorithms have been successfully applied to different problems such as requirement classification. Although these algorithms have good performance, most of them cannot explain how they make a decision. Explainable Artificial Intelligence (XAI) is a set of new techniques that explain the predictions of machine learning algorithms. In this work, the applicability of XAI for software requirement classification is studied. An explainable software requirement classifier is presented using the LIME algorithm. The explainability of the proposed method is studied by applying it to the PROMISE software requirement dataset. The results show that XAI can help the analyst or requirement specifier to better understand why a specific requirement is classified as functional or non-functional. The important keywords for such decisions are identified and analyzed in detail. The experimental study shows that the XAI can be used to help analysts and requirement specifiers to better understand the predictions of the classifiers for categorizing software requirements. Also, the effect of the XAI on feature reduction is analyzed. The results showed that the XAI model has a positive role in feature analysis.

Motivation & Objective

• To investigate the applicability of Explainable AI (XAI) techniques in software requirement classification.
• To develop an explainable classifier for distinguishing between functional and non-functional requirements.
• To evaluate how XAI improves analyst understanding of classifier predictions in requirement categorization.
• To analyze the impact of XAI on feature selection and reduction in requirement classification tasks.
• To identify key textual features that influence classification decisions using explainability techniques.

Proposed method

• The study employs the LIME (Local Interpretable Model-agnostic Explanations) algorithm to generate local explanations for predictions made by a machine learning classifier on software requirements.
• A supervised learning model is trained on the PROMISE software requirement dataset to classify requirements as functional or non-functional.
• LIME is applied post-training to interpret individual predictions by perturbing input features and observing prediction changes locally.
• Explanations are generated in the form of feature importance scores, highlighting which words or phrases contributed most to the classification outcome.
• The method enables analysts to inspect why a specific requirement was classified in a certain way, supporting traceability and validation.
• Feature reduction is analyzed by assessing the frequency and relevance of high-impact features identified by LIME across multiple predictions.

Experimental results

Research questions

• RQ1 To what extent can XAI techniques like LIME improve the interpretability of machine learning models in software requirement classification?
• RQ2 Which textual features are most influential in determining whether a requirement is classified as functional or non-functional?
• RQ3 How does the use of XAI affect the process of feature selection and reduction in requirement classification?
• RQ4 Can XAI help requirement engineers better understand and validate classifier decisions?
• RQ5 What is the consistency and reliability of LIME-generated explanations across diverse requirement instances?

Key findings

• The XAI-enhanced classifier significantly improves transparency by providing human-readable explanations for each prediction, enabling analysts to validate decisions.
• Key words such as 'should', 'must', 'user', and 'performance' were consistently identified as major contributors to classification outcomes.
• LIME-generated explanations revealed that functional requirements often contain action verbs and specific conditions, while non-functional requirements emphasized quality attributes.
• The XAI model demonstrated a positive role in feature analysis, helping to identify and prioritize relevant features for classification.
• Feature reduction was supported by XAI, as frequently occurring high-impact features were systematically identified and analyzed.
• The experimental evaluation confirmed that XAI enhances trust and usability of machine learning models in requirement engineering workflows.