Understanding the Issue Types in Open Source Blockchain-based Software Projects with the Transformer-based BERTopic
Md Nahidul Islam Opu, Md Shahidul Islam, Sara Rouhani, Shaiful Chowdhury · 2025
Read Understanding the Issue Types in Open Source Blockchain-based Software Projects with the Transformer-based BERTopic by Md Nahidul Islam Opu, Md Shahidul Islam, Sara Rouhani, Shaiful Chowdhury (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-11451.txt.
Introduction
Md Nahidul Islam Opu and co-authors mine issue reports from open source blockchain projects, in a 2025 paper. The head gives a shorter first name for the second author than the catalog. Blockchain systems are now used in many fields, yet their development pains were not mapped in a systematic way.
The team mined close to half a million issues from more than twelve hundred GitHub projects. A transformer based topic tool sorted them into forty nine topics in eleven groups. General software issues and blockchain specific ones split the set near evenly. Wallet management was the top topic and the slowest to resolve, while mechanism issues closed fastest.
Catalog record
| CREATOR | Md Nahidul Islam Opu, Md Shahidul Islam, Sara Rouhani, Shaiful Chowdhury |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-11451
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Understanding the Issue Types in Open Source Blockchain-based Software Projects with the Transformer-based BERTopic Md Nahidul Islam Opu, Md Shahidul Islam, Sara Rouhani, Shaiful Chowdhury Abstract Blockchain-based software systems are increasingly deployed across diverse domains, yet a systematic understanding of their development challenges remains limited. This paper presents a large-scale empirical study of 497,742 issues mined from 1,209 open-source blockchain projects hosted on GitHub. Employing BERTopic, a transformer-based topic modeling technique, we identify 49 distinct issue topics and organize them hierarchically into 11 major subcategories. Our analysis reveals that both general software development issues and blockchain-specific concerns are nearly equally represented, with Wallet Management and UI Enhancement emerging as the most prominent topics. We further examine the temporal evolution of issue categories and resolution times, finding that Wallet issues not only dominate in frequency but also exhibit the longest resolution time. Conversely, Mechanisms issues are resolved significantly faster. Issue frequency surged after 2016 with the rise of Ethereum and decentralized applications, but declined after 2022. These findings enhance our understanding of blockchain software maintenance, informing the development of specialized tools and practices to improve robustness and maintainability. I Introduction Blockchain technology has emerged as a transformative innovation with the potential to revolutionize a wide range of sectors [ 1 ] by offering enhanced transparency, security, and immutability. Beyond its foundational role in powering cryptocurrencies, blockchain is increasingly being adopted across diverse domains such as supply chain management [ 2 , 3 ] , healthcare [ 4 ] , and finance [ 5 ] . The rapid growth of the cryptocurrency market, which has surpassed a trillion-dollar valuation [ 6 ] , has further accelerated interest and investment in blockchain-based solutions. This widespread adoption has led to a surge in the development of blockchain application projects, creating a vibrant and rapidly evolving ecosystem of blockchain software development. Developing high-quality and maintainable applications requires a clear understanding of the nature and characteristics of software issues [ 7 , 8 ] encountered during development, testing, and maintenance. These characteristics include the distribution of issues across different components [ 9 ] , their impact on the system, resolution times, and the correlation between various issue types. Gaining insights into these aspects can guide the prioritization of development and testing [ 10 ] efforts. For instance, if a significant number of issues arise from cryptographic security flaws, it would indicate the need for more specialized tools, deeper domain expertise, and a stronger emphasis on security-focused development practices. Such understanding enables more informed decision-making and contributes to the creation of more robust, secure, and maintainable applications. Recent research has focused on identifying issues and bugs in blockchain-based systems, aiming to uncover the underlying characteristics and trends that developers encounter. However, these studies have either focused narrowly on bugs within a limited number of blockchain projects [ 11 ] or examined other types of development activities, such as commits and pull requests [ 12 ] . To the best of our knowledge, no prior research has undertaken a systematic, large-scale analysis of issue reports from blockchain-based applications on GitHub to comprehensively characterize the development challenges prevalent in this domain. To address this gap, we collected 497,742 issues from a curated set of 1,209 open-source blockchain-based projects on GitHub and applied topic modeling techniques to uncover the key development challenges faced by practitioners in this domain. Through this large-scale empirical analysis, we aim to answer the following three research questions: RQ1: What are the key issue categories in blockchain projects as identified through topic analysis? By leveraging advanced topic modeling techniques using BERTopic on GitHub issues, followed by systematic manual annotation, we identified and categorized 49 distinct topics across the analyzed projects. Among these, the most dominant were Wallet Management and Connectivity (21.3%), UI Enhancement (10.6%), and Logging and Error Handling (9.2%). RQ2: How do resolution times vary across different subcategories of issues in blockchain projects? The resolution time analysis shows that Wallet issues take notably longer to resolve, whereas Mechanisms issues, such as those involving consensus protocols, the Network Nervous System (NNS), and neuron staking, are resolved more quickly. RQ3: How have different types of issues evolved over time in terms of frequency within blockchain projects? We examined the temporal trends in issue creation across key subcategories and identified a significant surge in reported issues beginning around 2016. This surge was particularly pronounced in categories such as Development & Maintenance , User Experience & Interaction , and Wallet . To facilitate replication and extension, we publicly share our data and code 1 1 1 https://github.com/SQMLab/StudyOnBlockchainApps . II Background & Related Works Numerous studies have explored issues, particularly from GitHub, to understand developer concerns, issue trends, and resolution patterns. Tamanna et al. [ 13 ] conducted a large-scale empirical study of 118K GitHub issues from 34 runtime system repositories to categorize issue types and analyze issue management practices. Beyer and Pinzger [ 14 ] conducted a manual classification of Android development issues on Stack Overflow, uncovering recurring challenges faced by mobile developers. In addition to manual categorization, various automated approaches have been used to analyze issue-fixing times, particularly for bugs. For example, Lamkanfi et al. [ 15 ] observed skewed bug-fix time distributions in Eclipse and Mozilla, while Panjer [ 16 ] applied logistic regression to predict bug resolution times using Eclipse Bugzilla data. Giger et al. [ 17 ] advanced this line of work using decision-tree-based models, achieving improved precision and recall. More recently, Ardimento et al. [ 18 ] applied BERT-based models to predict bug-fixing durations, reporting accuracy up to 91%. Topic modeling has been extensively applied to mine latent themes in developer communications, including GitHub, Stack Overflow, and issue trackers. Thung et al. [ 8 ] and Uddin et al. [ 9 ] used Latent Dirichlet Allocation (LDA) to identify 40 prevalent topics in bug reports and developer Q&A, respectively. Bangash et al. [ 19 ] explored how developer discussions evolve temporally in the machine learning domain on Stack Overflow, revealing trends and tagging challenges by identifying 50 topics. More recently, BERTopic has emerged as a state-of-the-art topic modeling method that combines transformer-based embeddings with clustering techniques to extract coherent and semantically meaningful topics. Grootendorst [ 20 ] demonstrated that BERTopic outperforms classical models in coherence and human evaluation. Egger et al. [ 21 ] found BERTopic particularly effective for modeling short and noisy texts, such as Twitter data, demonstrating its versatility. Several recent studies have applied topic modeling with GitHub issue analysis to study blockchain-related software development. Vaccargiu et al. [ 22 ] utilized BERTopic on GitHub issues and comments to uncover sustainability-related themes in Ethereum developer discussions, focusing on the Go-Ethereum repository. In a related study, Vaccargiu and Tonelli [ 23 ] applied BERTopic to blockchain projects in the environmental sector, identifying core concerns such as carbon credit tracking and resource optimization. Wan et al. [ 11 ] manually classified 946 unique bugs across eight blockchain systems into ten categories using card sorting. They also investigated bug-fixing dynamics in blockchain systems, revealing that security bugs took the longest, while performance bugs often remained unresolved for extended periods. Despite extensive work on issue analysis, topic modeling, and blockchain software, there remains a gap in