IEEE Transactions on Software Engineering最新文献

{"title":"Text-to-Mapper: Generating Executable MyBatis ORM Artifacts From Natural Language Requirements","authors":"Zhenhan Chen;Xiaoyu Lin;Yingxin Chen;Ruoyu Huang;Jiqiang Yan;Zhenghao Chi;Jiahao Dai;Zhiling Cai;Lijin Wang;Yiwen Zhong","doi":"10.1109/TSE.2026.3702824","DOIUrl":"https://doi.org/10.1109/TSE.2026.3702824","url":null,"abstract":"Enterprise database applications are typically mediated by Object–Relational Mapping (ORM) frameworks such as MyBatis rather than by raw SQL, yet most natural-language interfaces to data target single SQL strings and bypass the mapper layer. This paper formulates the pioneering task of <italic>Text-to-Mapper</i>, generating executable MyBatis mapper files from natural-language requirements given database schemas and Java entity specifications. To support this setting, a high-quality dataset, <italic>MyBatis-Spider</i> together with an evaluation pipeline that exercises candidates through a parse-build-execute-align runtime chain. To capture execution-level semantics, an execution-based metric, <italic>Object Semantic Alignment</i> (OSA), compares the Java objects produced by generated and reference mappers. Our empirical study of direct-transfer baselines highlights recurring field-column-type binding errors; motivated by this, we introduce <italic>Across-Struct-Table (AST)</i>, a compact two-dimensional representation that makes these correspondences explicit. Systematic experiments with commercial and open-weights large language models show that AST yields consistent OSA improvements by reducing field-level mismatches. With parameter-efficient fine-tuning (PEFT) on top of AST, mid-scale open-weights models narrow the gap to larger general-purpose models, and in some settings match or slightly exceed them on this benchmark, suggesting a viable path toward controllable, on-premise Text-to-Mapper systems.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2688-2703"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148985843","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Boosting High-Review-Value Warning Line Identification With Unsupervised Line-Level Defect Prediction","authors":"Yufei Zhou;Hao Ding;Haihua Tang;Longtao Zhu;Zhide Zhou;Junyan Qian","doi":"10.1109/TSE.2026.3704180","DOIUrl":"10.1109/TSE.2026.3704180","url":null,"abstract":"Static analysis tools (SATs) are essential for detecting potential defects in software, yet their practical value is often undermined by the sheer volume of warning lines they produce. This flood of warning lines makes manual inspection prohibitively expensive and inefficient, especially when developers cannot readily discern which warning lines carry high review value, those most likely to lead developers to discover actual defects. Current priority-based prioritization schemes focus predominantly on the warnings themselves, neglecting a crucial factor: the intrinsic defect-proneness of the underlying code lines. Without quantifying such latent risks, it remains difficult to distinguish which warning lines deserve immediate attention. To bridge this gap, we propose APART, a lightweight framework that boosts the identification of high-review-value warning lines by integrating unsupervised line-level defect prediction (LDP). APART integrates unsupervised LDP prediction (reflecting structural risks at the line level) with warning priority and location data to construct a multi-level ranking. This enables developers to systematically prioritize warning lines that are not only severe but also located in code regions intrinsically prone to defects. A key strength of APART is its unsupervised and history-free design: it requires no labeled defect data, version history, or project-specific training, ensuring immediate applicability across diverse projects and SATs. We evaluated APART on nine real-world projects using eight popular static analysis tools. The results show that APART substantially outperforms the native priority-based baseline and stands competitive with representative LDP methods. Remarkably, the APART-F variant achieves top-tier performance across six key evaluation metrics, surpassing all existing LDP methods. By leveraging unsupervised LDP to boost identification accuracy at zero labeling cost, APART establishes a robust benchmark for future research and empowers developers to focus their efforts where it matters most.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2821-2837"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148288627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ICE: Reducing Search Space for Error-Inducing Input Detection","authors":"Youshuai Tan;Zhanwei Zhang;Zishuo Ding;Jinfu Chen;Weiyi Shang","doi":"10.1109/TSE.2026.3703362","DOIUrl":"10.1109/TSE.2026.3703362","url":null,"abstract":"Floating-point programs play a crucial role in scientific and engineering applications as well as daily life. Due to the inherent inaccuracies of floating-point arithmetic, these programs are susceptible to errors. While only a small subset of possible inputs may lead to significant inaccuracies in the program’s final results, such errors can have serious consequences. Prior approaches have been developed to identify error-inducing inputs using search algorithms. However, the task is challenging due to the sparsity of the search space. Furthermore, these tools frequently depend on high-precision programs as oracles for computing floating-point errors. This reliance introduces additional challenges, including prolonged execution times and instability. To detect these inputs more effectively and efficiently, we propose a simple yet effective method called ICE, designed to generate reduced and erroneous domains for these approaches. ICE first calculates the errors of the target program by taking all half-precision floating-point numbers as inputs. Instead of calculating errors based on high-precision programs, ICE injects small perturbations into each atomic operation of the program and measures the difference between the perturbed and original results to estimate floating-point errors. Subsequently, ICE selects the numbers with errors in the top <inline-formula><tex-math>$boldsymbol{n%}$</tex-math></inline-formula> as the endpoints of the important ranges. These individual ranges, which form domains, can then be used to improve existing tools by reducing the search space. To evaluate ICE, we apply ICE with three state-of-the-art error-inducing input detection approaches, i.e., ATOMU, HSED, and FPCC. Experimental results demonstrate that ICE could improve the performance of existing approaches by effectively reducing the search space, detecting more error-inducing inputs, and significantly shortening the search time.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2760-2773"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148288634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"HF-DGF: Hybrid Feedback Guided Directed Grey-Box Fuzzing","authors":"Guangfa Lyu;Zhenzhong Cao;Xiaofei Ren;Fengyu Wang","doi":"10.1109/TSE.2026.3701683","DOIUrl":"10.1109/TSE.2026.3701683","url":null,"abstract":"Directed Grey-box Fuzzing (DGF) is widely used for crash reproduction and patch testing because it can steer execution toward predefined target locations. However, existing DGF tools often rely on limited runtime feedback, which weakens both target convergence and target-state exploration. This paper presents <b> <small>HF-DGF</small> </b>, a directed grey-box fuzzing framework with hybrid feedback-guided seed scheduling. <b> <small>HF-DGF</small> </b> integrates control-flow distance, value-flow influence score, and slice coverage. For precise distance guidance, we design a backward-stepping algorithm that computes basic-block-level distances on a virtual interprocedural control-flow graph (ICFG). For deeper target-state exploration, we introduce value-flow influence and its corresponding metric, the value-flow influence score. To control the runtime cost introduced by hybrid feedback, we further adopt a selective instrumentation strategy. Evaluation on 41 real-world vulnerabilities (32 valid for comparative analysis) shows that <b> <small>HF-DGF</small> </b> consistently outperforms existing tools. On average, it reproduces crashes 5.05<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than AFL, 50.56<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than AFL<inline-formula> <tex-math>${+}{+}$</tex-math> </inline-formula>, 5.79<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than AFLGo, 73.75<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than WindRanger, 2.56<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than DAFL, and 8.45<inline-formula> <tex-math>$mathbf{times}$</tex-math> </inline-formula> faster than Beacon. Notably, among cases where all fuzzers trigger the target crash, <b> <small>HF-DGF</small> </b> achieves the lowest code coverage, indicating stronger directionality. It also outperforms AFLGo, WindRanger, DAFL, and Beacon in static-analysis efficiency.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2638-2653"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148288677","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards Effective Sampling of SMT Solutions","authors":"Shuangyu Lyu;Chuan Luo;Ruizhi Shi;Wei Wu;Chanjuan Liu;Chunming Hu","doi":"10.1109/TSE.2026.3702537","DOIUrl":"https://doi.org/10.1109/TSE.2026.3702537","url":null,"abstract":"This work focuses on effectively generating diverse solutions for satisfiability modulo theories (SMT) formulas, targeting the theories of bit-vectors, arrays, and uninterpreted functions, which is a critical task in software and hardware testing. Generating diverse SMT solutions helps uncover faults and detect safety violations during verification and testing. This gives rise to the SMT sampling problem, whose goal is to construct a diverse solution set that achieves comprehensive coverage of the constraint space. While high coverage is crucial for exploring system behaviors, the number of generated solutions should also be kept small in software engineering scenarios. Since each SMT solution typically corresponds to a test case or test input, generating fewer solutions for the same coverage directly reduces testing time and resource consumption, thereby improving testing efficiency. In this work, we introduce <italic>PanSampler</i>, a novel SMT sampler that achieves high coverage with a small number of solutions. <italic>PanSampler</i> incorporates a new diversity-aware SMT algorithm and a post-sampling optimization strategy, which together enhance its practical performance. It iteratively samples solutions, evaluates candidates, and employs local search to refine solutions, ensuring high coverage with a small number of samples. Extensive experiments on practical benchmarks demonstrate that <italic>PanSampler</i> exhibits a significantly stronger capability to reach high target coverage, while requiring fewer solutions than state-of-the-art samplers to achieve the same coverage level. Furthermore, our empirical evaluation on practical subjects, which are collected from real-world software systems, shows that <italic>PanSampler</i> achieves stronger fault detection capability. Moreover, when reaching the same fault detection effectiveness as the competitors, <italic>PanSampler</i> greatly reduces the number of required solutions, thereby improving testing efficiency. Our evaluations also validate the effectiveness of <italic>PanSampler</i>’s core components. <italic>PanSampler</i> represents a significant advancement in SMT sampling for software engineering, contributing to the cost reduction of software testing and hardware verification.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2670-2687"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11557498","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148985352","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"No Resource, No Benchmarks, No Problem? Evaluating and Improving LLMs for Code Generation in No-Resource Languages","authors":"Alessandro Giagnorio;Alberto Martin-Lopez;Gabriele Bavota","doi":"10.1109/TSE.2026.3703553","DOIUrl":"10.1109/TSE.2026.3703553","url":null,"abstract":"Large Language Models (LLMs) have significantly advanced the automation of software engineering tasks. One prominent example is code generation, where an LLM produces code in a specified programming language based on a natural language description. Most research in this area has focused on high-resource languages, such as Python or Java, which benefit from abundant training data in repository platforms like GitHub. A smaller body of work has explored low-resource languages (<italic>e.g.,</i> Lua, Racket), which are underrepresented in training corpora. In contrast, no-resource languages for which LLMs have seen virtually no training data remain largely unstudied. These languages often emerge in industry, where organizations develop proprietary or domain-specific languages unsupported by commercial tools like GitHub Copilot. This results in the need for companies to deploy their own in-house code recommenders. To investigate possible solutions in this context, we build and release three code generation benchmarks for no-resource languages, based on two recently proposed programming languages for which very little training data is available. Using these benchmarks, we experiment several solutions to teach LLMs about no-resource languages, including prompt-based techniques (<italic>e.g.,</i> few-shot) as well as pre-training and fine-tuning exploiting the little data available. While further pre-training gives the largest performance gains for no-resource languages, applying it directly to instruction-tuned models harms their ability to follow instructions. To address this, we start from a base model, further pre-training it on the target language, and then inject instruction-following capabilities via weight diff transfer from an instruction model. Such an approach significantly improves code generation capabilities in no-resource settings, allowing companies to cheaply deploy an instruct model specialized on the language of interest without dealing with the computational cost of instruction fine-tuning.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2774-2788"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11563597","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148288635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Characterizing Tests in IoT Software: Practices, Challenges and Opportunities","authors":"Rufeng Chen;Hengcheng Zhu;Wuqi Zhang;Zixu Zhou;Lili Wei","doi":"10.1109/TSE.2026.3700841","DOIUrl":"10.1109/TSE.2026.3700841","url":null,"abstract":"The Internet of Things (IoT) is experiencing rapid growth. Smart devices are emerging in smart homes and industrial applications, performing mission-critical tasks. Bugs in IoT software can lead to severe consequences. For example, a buggy smart lock can allow unauthorized access to a private property. Testing is a primary practice to expose software bugs and ensure software quality. However, little is known about how IoT software is tested. To bridge this gap, we conducted the first empirical study on test cases in open-source IoT software. Specifically, we evaluated the effectiveness of test cases in IoT software, explored the challenges inherent in testing IoT software, and analyzed the usage of mock objects. Our results indicate that while IoT software often contains a considerable number of tests, their effectiveness remains limited. We identified the primary challenges in testing IoT software as managing complex interactions with various external dependencies, such as other network-reliant IoT components, file systems, operating systems, and databases. We also observed that the use of mock objects in IoT software closely aligns with our identified testing challenges. This alignment demonstrates the potential of mocking as a solution to enhance test coverage and address the complexities of IoT software testing.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2587-2602"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148288705","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Process Fragment Recommendation in Process Modeling: Are LLMs the Answer?","authors":"Mahdi Saeedi Nikoo;Sangeeth Kochanthara;Önder Babur;Mark van den Brand","doi":"10.1109/TSE.2026.3690186","DOIUrl":"10.1109/TSE.2026.3690186","url":null,"abstract":"A process model is a structured representation that captures sequence of activities and their relationships that form a process. Process models help to understand, analyze, and potentially improve a process by highlighting key steps, dependencies, and potential bottlenecks. Within software engineering, process models are increasingly used to represent system behavior, support requirements engineering, facilitate stakeholder communication, and guide the design of (enterprise) systems. Creating, maintaining, and extending these models remains a challenge, as it requires modeling expertise. Recommendation of potential model completions during model development can help avoid errors, streamline workflows, enhance efficiency, and accelerate model development. This study investigates the potential of using traditional process model similarity measuring tools and large language models (LLMs) as recommenders for process model completion. We compare the effectiveness of two most prominent similarity tools and two state-of-the-art LLMS as process completion recommenders. Our results indicate that LLMs might not be the answer always, with process model similarity measuring tools outperforming LLMs in process recommendations for processes with a maximum of 6 nodes, with three times more meaningful recommendation cases. Our results also reveal that both LLMs and similarity tools perform most reliably when <inline-formula><tex-math>$approx$</tex-math></inline-formula>60% of the complete process is given as input. This study highlights the importance of aligning recommendation techniques with available context, process size, and dataset characteristics.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2566-2586"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148291603","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"HiSDevRec: Hybrid Interaction and Semantic-Aware Software Developer Recommendation","authors":"Xinqiang Xie;Jiahui Liu;Xiaochun Yang;Bin Wang","doi":"10.1109/TSE.2026.3703189","DOIUrl":"https://doi.org/10.1109/TSE.2026.3703189","url":null,"abstract":"Software developer recommendation is a critical task for online collaboration platforms. Identifying suitable developers remains challenging due to difficulties in capturing interaction behavioral preferences and semantic relevance, as well as handling data sparsity and cold-start issues. Traditional approaches typically model developers’ competencies (e.g., skills, expertise, and experience) or infer their preferences from historical task interactions. However, they often underperform because they fail to address several key challenges: (1) effectively capturing deep interaction behavioral preferences; (2) modeling latent textual semantics between developers and tasks; and (3) integrating behavioral preferences with semantic relevance in a unified manner. To address these challenges, we propose HiSDevRec, a novel dual-modality hybrid framework for developer recommendation that jointly models interaction-based behavioral preferences and textual semantic relevance. Specifically, we introduce a hybrid representation learning approach that integrates ID-based behavioral patterns with semantic representations derived from textual information. We further employ graph attention networks to capture higher-order propagation over the interaction graph, generating enriched representations for developers and tasks. In addition, we design a self-supervised disentanglement mechanism to explicitly separate behavioral and semantic representations, improving recommendation performance while enabling factor-level interpretability. Finally, we conduct extensive experiments on two real-world datasets, along with a case study and a user survey in an industrial setting. The results demonstrate that HiSDevRec consistently outperforms state-of-the-art baselines.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2740-2759"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148985842","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards Robust Detection for Malicious Injection Variants","authors":"Yanjing Yang;He Zhang;Bohan Liu;Jinwei Xu;Xin Zhou;Jun Lyu;Zhewen Mao;Dongxue Pan","doi":"10.1109/TSE.2026.3656977","DOIUrl":"10.1109/TSE.2026.3656977","url":null,"abstract":"Web applications deployed in cloud platforms expose APIs that are vulnerable to injection attacks. Cloud platform providers need to establish Web Application Firewalls (WAFs) at application API gateways to mitigate these risks. However, WAFs often struggle to intercept Malicious Injection Variants (MIVs) generated through mutations that differ from historical attack patterns. Adversarial training, which involves generating MIVs using techniques similar to those used by attackers and then refining detection models in WAFs with these MIVs, is a promising means of enhancing detection capabilities. Driven by this challenge, we propose AdaRode, which is an adversarially augmented robust detector training method designed from the perspective of defenders. To ensure the diversity of the generated MIVs, we comprehensively collected mutation meta-rule for SQL and XSS injections. To improve the efficiency of generating MIVs, we propose a sampling algorithm that uses internal model outputs as feedback for validating generated variants. Additionally, we conducted experiments to select the suitable model backbone. The evaluation results demonstrate that the detector generated by AdaRode significantly outperforms existing baseline methods in detecting both original attacks and attacks using MIVs. Our method improves the detection of MIVs by 10% of <inline-formula><tex-math>$F_{1}$</tex-math></inline-formula> compared with baseline detectors. For generating MIVs, the adversarial algorithm within our method significantly outperforms the existing algorithms regarding the comprehensive consideration of generation efficiency and diversity.","PeriodicalId":13324,"journal":{"name":"IEEE Transactions on Software Engineering","volume":"52 9","pages":"2528-2545"},"PeriodicalIF":6.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146042618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}