Asli Sari
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Asli Sari has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Ayse Tosun
Faculty of Computer and Informatics Engineering, Istanbul Technical University, Istanbul, Trkiye
2 shared publications - Gülfem Işıklar Alptekin
Galatasaray University
2 shared publications Gulfem Isiklar Alptekin
Department of Computer Engineering, Galatasaray University, Istanbul, Trkiye
1 shared publication
Latest Publications
- Open Access
- 8 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 166-167
Automated Annotation of Test Types using LLMs: A Multi-LLM Workflow for Machine Learning Test-Related Commits
Test-type annotation classifies commits based on the type of software test introduced or modified. While there are established taxonomies for testing machine learning (ML) software systems, implementing these manually at the commit level can be costly. This makes large language models (LLMs) a promising alternative. Utilizing an eight-category test-type taxonomy, we evaluate 97 manually labeled commits from scikit-learn using GPT-4o, Claude Sonnet, and Qwen 2.5-7B-Instruct-Turbo LLMs through four prompting strategies. This analysis focused on their classification performance, stability, agreement with experts, top-2 candidate recovery, and multi-LLM voting workflows (S1-S4) along with a corresponding analysis on Keras. Notably, GPT-4o in a zero-shot setting demonstrated the highest stability, and permitting a second label led to a reduction in single-prediction errors (e.g., Claude CoT-ZS Hit@1 = [59.79]%, Hit@2 = [77.32]%). Our most effective voting strategy (S1) surpassed the solo baseline, with its superior macro-F1 score (S1 > S2 > S3 > S4) also holding for Keras.
- 8 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 166-167
Open AccessAutomated Annotation of Test Types using LLMs: A Multi-LLM Workflow for Machine Learning Test-Related Commits
Test-type annotation classifies commits based on the type of software test introduced or modified. While there are established taxonomies for testing machine learning (ML) software systems, implementing these manually at the commit level can be costly. This makes large language models (LLMs) a promising alternative. Utilizing an eight-category test-type taxonomy, we evaluate 97 manually labeled commits from scikit-learn using GPT-4o, Claude Sonnet, and Qwen 2.5-7B-Instruct-Turbo LLMs through four prompting strategies. This analysis focused on their classification performance, stability, agreement with experts, top-2 candidate recovery, and multi-LLM voting workflows (S1-S4) along with a corresponding analysis on Keras. Notably, GPT-4o in a zero-shot setting demonstrated the highest stability, and permitting a second label led to a reduction in single-prediction errors (e.g., Claude CoT-ZS Hit@1 = [59.79]%, Hit@2 = [77.32]%). Our most effective voting strategy (S1) surpassed the solo baseline, with its superior macro-F1 score (S1 > S2 > S3 > S4) also holding for Keras.
- Open Access
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 83784-83825
A Systematic Mapping Study on Developer Profile Modeling in Software Engineering
This study addresses the critical role of developer profile modeling in today’s dynamic software industry. Following mapping study guidelines, we reviewed 254 primary studies from five databases, and analyzed them against five research questions to report profile definitions, relevant software engineering roles, data sources, measures to quantify the profiles, and tasks used for empirical analysis. Our analysis highlights the need for standardized profile definitions due to many characteristics, such as gender, expertise, behavior, activity, chosen in the literature. The most frequently analyzed people are software developers, particularly for their role in coding. A notable trend is using profile modeling to characterize developers, focusing on personality traits, gender, and skills, and through mining data from GitHub. We also identified an inevitable research direction that revolves around developer profiling, particularly regarding interactions with Large Language Model agents. This sharing behavior can foster diverse profile characteristics and lead to the development of quantifiable metrics for profile definitions in artificial intelligence-assisted contexts. We present a mapping between profile definitions, data types (industry versus academia), metric groups to quantify these definitions, and their applications in two software engineering tasks. Our analysis suggests the need for future studies to define more generic profile modeling through multiple characteristics, better quantification of these characteristics through independent measures, and more analysis on different software engineer roles, such as testers. Exploring the sharing behaviors associated with artificial intelligence and Large Language Model agents will significantly enhance developer profiling by promoting the adoption of artificial intelligence-assisted software development.
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 83784-83825
Open AccessA Systematic Mapping Study on Developer Profile Modeling in Software Engineering
This study addresses the critical role of developer profile modeling in today’s dynamic software industry. Following mapping study guidelines, we reviewed 254 primary studies from five databases, and analyzed them against five research questions to report profile definitions, relevant software engineering roles, data sources, measures to quantify the profiles, and tasks used for empirical analysis. Our analysis highlights the need for standardized profile definitions due to many characteristics, such as gender, expertise, behavior, activity, chosen in the literature. The most frequently analyzed people are software developers, particularly for their role in coding. A notable trend is using profile modeling to characterize developers, focusing on personality traits, gender, and skills, and through mining data from GitHub. We also identified an inevitable research direction that revolves around developer profiling, particularly regarding interactions with Large Language Model agents. This sharing behavior can foster diverse profile characteristics and lead to the development of quantifiable metrics for profile definitions in artificial intelligence-assisted contexts. We present a mapping between profile definitions, data types (industry versus academia), metric groups to quantify these definitions, and their applications in two software engineering tasks. Our analysis suggests the need for future studies to define more generic profile modeling through multiple characteristics, better quantification of these characteristics through independent measures, and more analysis on different software engineer roles, such as testers. Exploring the sharing behaviors associated with artificial intelligence and Large Language Model agents will significantly enhance developer profiling by promoting the adoption of artificial intelligence-assisted software development.
- 1 December 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 450-453
A Prize Determination Approach for Crowdsourced Software Development
Although the dominance of in-house development, contracting and outsourcing, crowdsourcing has been accepted as an alternative software development approach in recent years. It is a form of collective intelligence to work on software engineering tasks, from requirement extraction to testing. In the most common crowdsourcing model, a client (a firm or some type of institution) first broadcasts the task/project, which is intended to be developed, via a crowdsourcing platform. Each task/project has a prize and a deadline. Then, members of the platform community/crowd/individuals select the tasks that they want to submit solutions, either individually or in a collaborative manner. The submitted solutions are evaluated by the crowdsourcing platform and the most successful one is remunerated by the initiating organization. The prize of each task may be determined in terms of the expected duration of the task, its complexity, its quality, the required effort to complete, and the online reputation of the client. TopCoder awarding mechanism, performance-based, game theory-based, deadline driven or revenue sharing models are several approaches for determining task prizes in crowdsourced software development (CSD). In this paper, Putnam model, a well-known software effort estimation model, is proposed to be applicable as a prize determination approach in crowdsourcing. The demonstrative examples are chosen from TopCoder projects, and Function Point Analysis is used to calculate expected effort of the projects.
- 1 December 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 450-453
A Prize Determination Approach for Crowdsourced Software Development
Although the dominance of in-house development, contracting and outsourcing, crowdsourcing has been accepted as an alternative software development approach in recent years. It is a form of collective intelligence to work on software engineering tasks, from requirement extraction to testing. In the most common crowdsourcing model, a client (a firm or some type of institution) first broadcasts the task/project, which is intended to be developed, via a crowdsourcing platform. Each task/project has a prize and a deadline. Then, members of the platform community/crowd/individuals select the tasks that they want to submit solutions, either individually or in a collaborative manner. The submitted solutions are evaluated by the crowdsourcing platform and the most successful one is remunerated by the initiating organization. The prize of each task may be determined in terms of the expected duration of the task, its complexity, its quality, the required effort to complete, and the online reputation of the client. TopCoder awarding mechanism, performance-based, game theory-based, deadline driven or revenue sharing models are several approaches for determining task prizes in crowdsourced software development (CSD). In this paper, Putnam model, a well-known software effort estimation model, is proposed to be applicable as a prize determination approach in crowdsourcing. The demonstrative examples are chosen from TopCoder projects, and Function Point Analysis is used to calculate expected effort of the projects.