Yoshiaki Fukazawa

The total number and diversity of smartphone users are increasing. Additionally, their usage varies such as communication, document processing and online shopping. Applications realize different usages to meet users’ requirements. However, some users are unfamiliar with smartphones. They have difficulty learning the operations and need support. A tutorial system, which provides operations along with users’ status, is effective to resolve these issues. However, showing both the target application’s screen and the tutorial on a smartphone display is challenging. Herein we propose a method to generate a system that shows tutorials on a Personal Computer (PC). The target application on a smartphone and the tutorial system on a PC are connected. The user’s status is sent to the tutorial system to realize a customized tutorial. To generate the tutorial system, a log obtaining function is added to the target application without modifying the source program. The log data of user’s operations are recorded. Operation flows in the target application are constructed and frequent operations are weighted. Then tutorials are generated and shown based on the weighting. Our method allows a user to view appropriate tutorials for the current status of the operation on a large display of a PC.

Resource-constrained environments, such as embedded devices, have limited amounts of memory and storage. Practical programming languages such as C++ and Rust tend to output multiple similar functions by monomorphizing polymorphic functions. An optimization technique called Function Merging, which merges similar functions into a single function, has been studied. However, in the state-of-the-art approach, the number of functions that can be merged at once is limited to two; thus, efficiently merging three or more functions, which are often generated from polymorphic functions, has been impossible. In this study, we propose Multiple Function Merging optimization, which targets merging three or more similar functions into a single function using a multiple sequence alignment algorithm. With multiple aligned information, Multiple Function Merging can increase merge opportunities and reduce extra branching overheads at the code generation stage. We evaluated it using the SPEC CPU benchmark suite and some large-scale C/C++ programs, and the results show that it reduces code size by as much as 7.61% compared with the state-of-the-art approach.

Advancements in large language models (LLMs) have enhanced their ability to handle ambiguous user instructions. However, effective prompt patterns remain crucial for usability and comprehension. This article presents a taxonomy of prompt engineering patterns for software engineering. It is based on a systematic literature review that was conducted in early 2023, when LLMs still faced significant limitations in context length and inference capabilities. Our study explores techniques that enhance the usability and reliability of LLMs, emphasizing the ongoing importance of well-designed prompts in optimizing task performance. Our findings highlight the critical role of prompt patterns in maximizing LLM’s potential, even as their capabilities continue to evolve.

Development of machine learning (ML) systems differs from traditional approaches. The probabilistic nature of ML leads to a more experimentative development approach, which often results in a disparity between the quality of ML models with other aspects such as business, safety, and the overall system architecture. Herein the Multi-view Modeling Framework for ML Systems (M³S) is proposed as a solution to this problem. M³S provides an analysis framework that integrates different views. It is supported by an integrated metamodel to ensure the connection and consistency between different models. To facilitate the experimentative nature of ML training, M³S provides an integrated platform between the modeling environment and the ML training pipeline. M³S is validated through a case study and a controlled experiment. M³S shows promise, but future research needs to confirm its generality.

In this study, we employ causal inference models, specifically Bayesian Networks (BN) and Linear Non-Gaussian Acyclic Models (LiNGAM), to investigate the determinants of Bug Fixing Time (BFT) in software development. Moving beyond traditional statistical analyses, our approach aims to identify the true causal factors influencing BFT. Our findings indicate that ’Reporter Reputation’, ’Severity’, and ’Blocker’ status are significant determinants of BFT, with notable differences between bugs reported by users versus developers. This research challenges existing assumptions about the necessity of comprehensive bug reports and underscores the importance of understanding bug resolution’s complexity and organizational context. By applying causal inference models, we offer actionable insights for improving bug prioritization, operational efficiency, and predictive management of development bottlenecks, enhancing the software development lifecycle. Our study bridges the theoretical and practical aspects of software quality optimization and introduces a novel perspective on managing software development processes. Additionally, our analysis reveals counterintuitive results that further contribute to our understanding of the dynamics influencing BFT.

Security-related issues have become more significant due to the proliferation of IT. Collating security-related information in a database improves security. For example, Common Vulnerabilities and Exposures (CVE) is a security knowledge repository containing descriptions of vulnerabilities about software or source code. Although the descriptions include various entities, there is not a uniform entity structure, making security analysis difficult using individual entities. Developing a consistent entity structure will enhance the security field. Herein we propose a method to automatically label select entities from CVE descriptions by applying the Named Entity Recognition (NER) technique. We manually labeled 3287 CVE descriptions and conducted experiments using a machine learning model called BERT to compare the proposed method to labeling with regular expressions. Machine learning using the proposed method significantly improves the labeling accuracy. It has an f1 score of about 0.93, precision of about 0.91, and recall of about 0.95, demonstrating that our method has potential to automatically label select entities from CVE descriptions.

Due to the criticality of software and IT in today’s business environment, organisations often align business goals and IT strategies. One alignment method is GQM+Strategies^® (GQM+Strategies^® is a registered trademark (No. 302008021763 at the German Patent and Trade Mark Office and international registration number IR992843). Although GQM+Strategies employs a vertical refinement tree grid based on rationales to align business goals and IT strategies in each department and throughout the organisation, it allows multiple perspectives. This can lead to strategic problems and risks because the GQM+Strategies grid may be unclear. To address this deficiency, this study defines modelling rules with a metamodel specified using a Unified Modelling Language (UML) class diagram and employs design principles described by Object Constraint Language (OCL) to automatically configure GQM+Strategies grids. These design principles are used in an experiment as the evaluation criteria to assess potential strategic problems and risks. The results confirm that our method has the potential to support the construction of GQM+Strategies grids with a consistent perspective, aiding in the alignment of business goals and IT strategies throughout an organisation.

Persona --- fictional user profiles --- are used to identify user requirements in software engineering. However, methods targeting revisions, especially for existing B2B services, remain sparse. This paper proposes a method that integrates several models, including k-means clustering, term frequency-inverse document frequency (TF-IDF), and generative AI. Users' behavior tendencies, pain points, and other attributes are output solely from clickstream log data, bypassing the traditional survey-based approaches of previous studies. Clickstreams are vectorized and categorized, whereas users are further analyzed on the basis of time and content of their clickstreams. A case study was conducted with evaluations carried out both quantitatively and qualitatively. The results suggest that, although some parameters still need improvement, fairly rated persona outcomes were attained.

Multi-View Modeling Framework for ML Systems (M³S) provides a framework to synchronize the experimental nature of machine learning and the deterministic side of traditional software engineering. However, understanding the framework's generality and limitations still requires further investigation. This paper compares the existing validation case study to a new case study of the OCT image diagnosis support system. The comparison between the two case studies shows M³S capability to handle variations in the nature of ML system analysis. However, the framework's capability to handle different ML tasks other than multi-class classification still requires further investigation.