International Journal of Software Engineering and Knowledge Engineering

With the advances in machine learning, there is a growing interest in AI-enabled tools for autocompleting source code. GitHub Copilot, also referred to as the “AI Pair Programmer”, has been trained on billions of lines of open source GitHub code, and is one of such tools that has been increasingly used since its launch in June 2021. However, little effort has been devoted to understanding the practices, challenges, and expected features of using Copilot in programming for auto-completed source code from the point of view of practitioners. To this end, we conducted an empirical study by collecting and analyzing the data from Stack Overflow (SO) and GitHub Discussions. More specifically, we searched and manually collected 303 SO posts and 927 GitHub discussions related to the usage of Copilot. We identified the programming languages, Integrated Development Environments (IDEs), technologies used with Copilot, functions implemented, benefits, limitations, and challenges when using Copilot. The results show that when practitioners use Copilot: (1) The major programming languages used with Copilot are JavaScript and Python, (2) the main IDE used with Copilot is Visual Studio Code, (3) the most common used technology with Copilot is Node.js, (4) the leading function implemented by Copilot is data processing, (5) the main purpose of users using Copilot is to help generate code, (6) the significant benefit of using Copilot is useful code generation, (7) the main limitation encountered by practitioners when using Copilot is difficulty of integration, and (8) the most common expected feature is that Copilot can be integrated with more IDEs. Our results suggest that using Copilot is like a double-edged sword, which requires developers to carefully consider various aspects when deciding whether or not to use it. Our study provides empirically grounded foundations that could inform software developers and practitioners, as well as provide a basis for future investigations on the role of Copilot as an AI pair programmer in software development.

Smart contracts are programs running on blockchain. In recent years, due to the persistent occurrence of security-related accidents in smart contracts, the effective detection of vulnerabilities in smart contracts has received extensive attention from researchers and engineers. Machine learning-based vulnerability detection techniques have the advantage that they do not need expert rules for determining vulnerabilities. However, existing approaches cannot identify vulnerabilities when the versions of smart contract compilers are updated. In this paper, we propose OC-Detector (Opcode Clustering Detector), a smart contract vulnerability detection approach based on clustering opcode instructions. OC-Detector learns the characteristics of opcode instructions to cluster them and replaces opcode instructions belonging to the same cluster with the ID of the cluster. After that, the similarity between the contract under analysis and contracts in the vulnerability database is calculated to identify vulnerabilities. The experimental results demonstrate that OC-Detector improves the F₁ value of detecting vulnerabilities from 0.04 to 0.40 compared to DC-Hunter, Securify, SmartCheck and Osiris. Additionally, compared to DC-Hunter, the F₁ value is improved by 0.27 when detecting vulnerabilities in smart contracts compiled by different versions of compilers.

With the development of network science, complex network analysis has received extensive attention. In recent years, influential spreaders identification has become a hot topic in the research of complex networks. In general, influential spreader identification algorithms are mainly divided into centrality-based algorithms and topology-based algorithms. However, centrality-based algorithms have to face the information limitation problem that leads to low accuracy for identifying influential spreaders. Topology-based algorithms have both structural and positional limitation problems that lead to low accuracy for identifying peripheral influential spreaders. In this paper, we focus on improving the situation and propose two influential spreader identification algorithms, NSC (neighborhood structure centrality) and NPC (neighborhood position centrality) from both the perspective of the centrality and the network topology. NSC algorithm collects various types of network structure information through structure embedding and clustering, so as to solve missing network structure information problem. NPC algorithm calculates neighborhood location information by improving the k-shell algorithm to tackle location limitation problem. Experimental results with fourteen baseline algorithms show that our proposed algorithms NSC and NPC can achieve higher accuracy.

The big data sampling method for real-time and high-speed streaming data is prone to lose the value and information of a large amount of discrete data, and it is not easy to make an efficient and accurate evaluation of the value characteristics of streaming data. The SDSLA sampling method based on mineral drilling exploration can evaluate the valuable information of streaming data containing many discrete data in real-time, but when the range of discrete data is irregular, it has low sampling accuracy for discrete data. Based on the SDSLA algorithm, we propose a dynamic drilling sampling method SDDS, which takes well as the analysis unit, dynamically changes the size and position of the well, and accurately locates the position and range of discrete data. A new model SDVEM is further proposed for data valuation, which evaluates the sample set from discrete, centralized, and overall dimensions. Experiments show that compared with the SDSLA algorithm, the sample sampled by the SDDS algorithm has higher evaluation accuracy, and the probability distribution of the sample is closer to the original streaming data, with the AOCV indicator being nearly 10% higher. In addition, the SDDS algorithm can achieve over 90% accuracy, recall, and F1 score for training and testing neural networks with small sampling rates, all of which are higher than the SDSLA algorithm. In summary, the SDDS algorithm not only accurately evaluates the value characteristics of streaming data but also facilitates the training of neural network models, which has important research significance in big data estimation.

Multi-behavior recommendation has gained significant attention in recent years for its ability to outperform single-behavior models. Current research related to multi-behavior models leaves room for improvement in the following two areas. First, the noise carried by individual behaviors and the additional noise generated during behavior processing is often overlooked, and these can ultimately degrade recommendation performance. Second, the specific time period of behavioral interactions and the frequency of interactions within that time period are also not taken into account. To address the above limitations, we propose a multi-behavior recommendation model integrating dynamic preferences (MB-DP) that captures dynamic interests while smoothing and denoising multi-behavior information. MB-DP extracts low and high-order semantics from various behaviors and unifies the measurements to generate interaction predictions. Additionally, it analyzes the interaction time and frequency of each behavior using gated recurrent units to capture the dynamic preferences of users and improve the prediction values. Extensive experimental results on two real-world datasets show that MB-DP significantly improves recommendation performance compared to the state-of-the-art baselines.

Code summarization is a task that aims at automatically producing descriptions of source code. Recently many deep-learning-based approaches have been proposed to generate accurate code summaries, among which pre-trained models (PTMs) for programming languages have achieved promising results. It is well known that source code written in programming languages is highly structured and unambiguous. Though previous work pre-trained the model with well-design tasks to learn universal representation from a large scale of data, they have not considered structure information during the fine-tuning stage. To make full use of both the pre-trained programming language model and the structure information of source code, we utilize Flow-Augmented Abstract Syntax Tree (FA-AST) of source code for structure information and propose GraphPLBART — Graph-augmented Programming Language and Bi-directional Auto-Regressive Transformer, which can effectively introduce structure information to a well PTM through a cross attention layer. Compared with the best-performing baselines, GraphPLBART still improves by 3.2%, 7.1%, and 1.2% in terms of BLEU, METEOR, and ROUGE-L, respectively, on Java dataset, and also improves by 4.0%, 6.3%, and 2.1% on Python dataset. Further experiment shows that the structure information from FA-AST has significant benefits for the performance of GraphPLBART. In addition, our meticulous manual evaluation experiment further reinforces the superiority of our proposed approach. This demonstrates its remarkable abstract quality and solidifies its position as a promising solution in the field of code summarization.

This research provides a comprehensive analysis of deontic expressions within the Code of Federal Regulations (CFR) Title 48, Federal Acquisition Regulations System, specifically focusing on obligations, permissions, prohibitions, and dispensations. Utilizing SHAMROQ, a systematic and rigorous methodology, the authors extract, classify, and analyze these expressions, quantify their prevalence, and identify common linguistic patterns within the legal text. The results show that obligations (71.3%) form most deontic expressions in CFR 48, indicating the heavily prescriptive nature of the document. Permissions also form a significant part (21.9%), suggesting the liberties and allowances are embedded within the regulatory framework. In contrast, prohibitions (5.4%) and dispensations (1.4%) are less frequent, indicating that the document leans more towards defining what is required or allowed rather than what is explicitly forbidden or exempted. This research also highlights the challenges encountered during the extraction process, providing insights into the complexities of parsing legal texts and the intricacies of deontic language. These challenges range from the technical difficulties of parsing a complex hierarchical document to the conceptual challenges of defining precise rulesets for regulations and provisions. In summary, the results deepen the understanding of regulatory compliance in software engineering and contribute to the development of more effective and efficient automated extraction tools.

Wireless sensing techniques for Human Activity Recognition (HAR) have been widely studied in recent years. At present, research on HAR based on Radio Frequency Identification (RFID) is changing from the tag attachment method to the tag non-attachment method. Affected by multipath, the current solutions in tag non-attachment scenarios mainly focus on single-object activity recognition, which is not suitable for multi-object scenarios. To address these issues, we propose STMultiple, a novel tag non-attachment activity recognition model for multi-object. The model first preprocesses the raw signal with filter and phase calibration, then it applies dilated convolution in the frequency domain to extract multi-object activity features, finally the feature pyramid structure and ProbSparse are used to optimize the vanilla Transformer-Encoder to enhance the activity recognition ability. Extensive experiments show that STMultiple can achieve recognition accuracy of up to 97.93% and down to about 90% in challenging environments ranging from two to five users, which has excellent performance compared to several state-of-the-art methods.

In mobile edge computing environment, intelligent inference services driven by DNN are highly sensitive to latency. Recently, collaborative inference between User Devices and Edge Servers (ESs) based on Deep Neural Networks (DNN) partition has achieved success in service acceleration. However, most of the existing collaborative acceleration schemes are partitioned for a single DNN inference task, which cannot quickly make partition decisions for a set of concurrent inference tasks, and often sacrifice inference accuracy. In addition, due to the limited resources of ESs, there is resource competition among concurrent requests, which makes the partitioned tasks cannot be offloaded to ESs in time for processing. Therefore, designing an efficient offloading scheme becomes essential. The task offloading schemes based on deep reinforcement learning can solve complex decision-making problems in high-dimensional state space, but they have problems such as insufficient sample diversity and easily falling into local optimum. In this paper, a Collaborative Inference Acceleration Scheme integrating DNN Partitioning and Task Offloading (CIAS-PnO) is proposed. First, while ensuring inference accuracy, the Collaborative DNN Layer Partitioning (CDLP) algorithm is designed with the goal of optimal latency. CDLP can reduce the problem scale of concurrent inference tasks partition by pruning operation and determine the partition decisions in time. Then, the Distributed Soft Actor-Critic (SAC)-based Partition Task Offloading algorithm (DSACO) is designed. DSACO supports SAC Agents to explore samples in parallel and share learning experiences, and uses the automatic entropy adjustment mechanism to improve the exploration efficiency of Agents, so as to avoid falling into local optimum and achieve efficient offloading of partition tasks. Experimental results on DNN benchmarks show that compared with the baseline acceleration schemes, CIAS-PnO achieves more than 19.8% acceleration performance improvement, and has higher convergence performance and task success rate.

Technical debt (TD) expresses the lack of internal quality directly affecting software evolution. Therefore, it has gained the attention of software researchers and practitioners recently. Software researchers have performed empirical studies to observe the perspective of TD in different software cultures and organizations. However, it is important to replicate such studies in more places and with more practitioners to strengthen the perception of TD. In this paper, we present the results of a set of new research questions from an evolved survey design of a survey replication in the Uruguayan software industry to characterize how the software industry professionals understand, perceive, and adopt TD management (TDM) activities. The results allow us to observe that different participant contexts (startups, government, job roles) show different levels of awareness and perception of TD. Details in the form of the adoption of each TDM activity were presented. We could observe some difficulties in conducting some TDM activities that the practitioners consider very important, especially in TDM and monitoring. Differences in specific organizational contexts like startups and government could indicate the need for research efforts in other software engineering communities that meet their specific TD challenges and needs.

Code generation has a positive significance in supporting software development, reducing labor intensity, and improving development efficiency. Some scholars use similar code information to enhance the quality of code generation. However, to improve the efficiency and accuracy of programming in daily development tasks, developers often search for similar samples as references. They get the code’s syntactic structure and semantic information from similar samples to assist in programming development. Inspired by this, we argue that similar samples are helpful for code generation. This paper proposes a CodeGen-Search model to improve code generation quality by incorporating similar samples. To fully utilize the information of similar samples, the model adopts the “pre-training $+$ fine-tuning” pattern. The model uses a minimum edit distance algorithm to find some similar samples with natural language (NL), and uses different encoders to extract the features of the NL and the code in similar samples. Experimental results show that our model efficiently improves the quality of the generated code. Compared to the state-of-the-art model, the CodeGen-Search model improves the BLEU by 1.5%, the Rough by 0.8% on the HS dataset, and the StrAcc by 0.5% on the ATIS dataset.

Published studies on agile effort estimation predominantly focus on comparisons of the accuracy of different estimation methods, while efficiency comparisons, i.e. how much time the estimation methods consume was not in the forefront. However, for practical use in software development, the time required can be a very important cost factor for enterprises, especially when the accuracy of different agile effort estimations is similar. In this study, we thus try to advance the current standard accuracy comparison between methods by introducing efficiency, i.e. time it takes to use a method as an additional dimension of comparison. We conduct this comparison between three agile effort estimation methods that were not yet compared in the literature, namely, Planning Poker, Bucket System and Affinity Estimation. For the comparison, we used eight student teams with 29 students who had to use all the effort estimation methods during the course where they had to finish a programming project in 3 weeks. The results indicate that after the students get used to using the different methods the accuracy between them is not statistically significantly different, however, the efficiency is. On average, Bucket System and Affinity Estimation methods take half as much time as Planning Poker.