Xi Guo

Semi-structured documents contain both hierarchical structure and field-level semantics. This makes similarity measurement difficult for methods that focus only on structure or only on global semantic representations. To address this problem, we propose a framework for schema-level semantic–structural similarity of semi-structured documents. We first introduce Semantic Tree Edit Distance (STED), which combines semantic node matching with hierarchical tree alignment. We then generate training pairs through controlled perturbations of real JSON documents and use their STED scores as automatic supervision. Based on this supervision, STED–GNN learns an efficient approximation of STED for repeated similarity computation. Experiments on MixedJSONBench-Doc show that STED–GNN achieves a mean ARI of 0.9799 for clustering, an accuracy of 0.9967 in the primary inductive classification evaluation, and a mean Recall@76 of 0.7289 on the low-similarity evaluation. After document representations are cached, STED–GNN reduces repeated pairwise scoring time by approximately 183.47 times relative to STED. These results show that jointly modeling semantic and structural information is effective for semi-structured document similarity, while learned approximation substantially reduces the cost of repeated pairwise comparison.

Large Language Models (LLMs) show considerable potential for materials-science question answering. However, LLM responses may still be affected by unsupported parametric associations, while dense Retrieval-Augmented Generation (RAG) can fragment relational evidence across text chunks. Moreover, general graph-based retrieval does not necessarily preserve the hierarchical relations and factual attributes required to resolve implicit material constraints. To address these limitations, we propose MCTD-KG, a multi-source heterogeneous knowledge graph integrated with a Knowledge-Enhanced RAG framework for complex material question answering. MCTD-KG adopts a Classification–Term–Data ontology to connect disciplinary taxonomies, domain concepts, semantic relations, and empirical records from toolbooks and the scientific literature. Through LLM-assisted knowledge extraction, entity normalization, and multi-source integration, the resulting graph contains more than 530,000 entities across three layers, including 61,768 text-extracted Term-layer entities. During inference, Dual-Channel Retrieval jointly retrieves query-relevant relational paths and associated material attributes, while an explicit semantic filtering stage screens candidate evidence against the query constraints. Evaluation on an expert-validated benchmark of 1577 questions shows that the proposed framework achieves an overall accuracy of 68.48%, compared with 17.40% for the zero-shot Pure LLM, 24.79% for the best Vanilla RAG setting, and 44.96% for GraphRAG. It also achieves 45.22% accuracy on four-hop questions, compared with 39.49% for GraphRAG. These results indicate that integrating multi-source domain knowledge with relation-preserved retrieval and attribute-supported filtering provides more focused and inspectable evidence, thereby supporting more accurate complex material question answering.

Ride-hailing platforms allow people to request rides, but sometimes they do not respond quickly enough. Due to the complex urban transportation system and the large geographical area of the city, the imbalance between supply and demand becomes a common problem. To solve this problem, many studies have proposed taxi repositioning methods. However, each of two mainstream methods (region-based and taxi-based) has its own advantages and disadvantages. In this paper, we propose the RSTR model, which is a two-layer model that takes advantages of the two mainstream methods. We model the problem as a partially observable Markov decision process and design the optimization objective. To generate more accurate repositioning strategies and improve model training outcomes, we propose a many-to-many scheduling mode and demonstrate its effectiveness. Extensive experiments on real-world datasets show that RSTR can effectively balance supply and demand and outperform other baseline methods.

A scalable, reusable, and broad-coverage unified material knowledge representation shows its importance and will bring great benefits to data sharing among materials communities. A knowledge graph (KG) for materials terminology, which is a formal collection of term entities and relationships, is conceptually important to achieve this goal. In this work, we propose a KG for materials terminology, named Materials Genome Engineering Database Knowledge Graph (MGED-KG), which is automatically constructed from text corpus via natural language processing. MGED-KG is the most comprehensive KG for materials terminology in both Chinese and English languages, consisting of 8,660 terms and their explanations. It encompasses 11 principal categories, such as Metals, Composites, Nanomaterials, each with two or three levels of subcategories, resulting in a total of 235 distinct category labels. For further application, a knowledge web system based on MGED-KG is developed and shows its great power in improving data sharing efficiency from the aspects of query expansion, term, and data recommendation.

In this paper, we propose the fine-grained geospatial knowledge graph (FineGeoKG), which can capture the neighboring relations between geospatial objects. We call such neighboring relations strong geospatial relations (SGRs) and define six types of SGRs. In FineGeoKG, the vertices (or entities) are geospatial objects. The edges (or relations) can have “sgr” labels together with properties, which are used to quantify SGRs in both topological and directional aspects. FineGeoKG is different from WorldKG, Yago2Geo, and other existing geospatial knowledge graphs, since its edges can capture the spatial coherence among geospatial objects. To construct FineGeoKG efficiently, the crucial problem is to find out SGRs. We improve the existing geospatial interlinking algorithm in order to find out SGRs faster. To answer SGR queries efficiently, we design an index to organize the SGR edges and improve the binary join method for subgraph matching. We conduct experiments on the real datasets and the experimental results show that the proposed algorithm is more efficient than the baseline algorithms. We also demonstrate the usefulness of FineGeoKG by presenting the results of complicated spatial queries which focus on structural and semantic information. Such queries can help researchers (for example, ecologists) find groups of objects following specific spatial patterns.

The road transportation of explosives is highly concerning due to its substantial impact on social safety. For the safety management of explosive transportation, e.g., transport route planning and emergency rescue, explosion consequence evaluation is of paramount importance. The consequence evaluation of explosion accidents is affected by many factors, especially spatial features, such as the location of transport vehicles, the distribution of buildings, and the presence of individuals around the road, etc. However, there is still a lack of quantification methods for building damage evaluation, human casualty evaluation that considers real-time population density, and efficient interactive damage evaluation methods. In this paper, we formalize three typical scenarios of damage evaluation for explosive road transportation accidents, i.e., explosion point-based, road segment-based, and route-based damage evaluation. For each scenario, we propose a Height-aware Hierarchical Building Damage (HHBD) model and a Shelter-aware Human Casualty (SHC) model for building damage evaluation and human casualty evaluation, respectively. We also develop a GIS-based interactive visualization platform that integrates multi-source geospatial data and that enables efficient geospatial computation. In addition, a case study of liquefied natural gas (LNG) transportation in Wuhan is demonstrated in order to verify the effectiveness and efficiency of the proposed system. The research results can support the decision-making process of explosive transportation safety warnings and emergency rescue.

Deep neural networks (DNNs) are increasingly being used in malware detection and their robustness has been widely discussed. Conventionally, the development of an adversarial example generation scheme for DNNs involves either detailed knowledge concerning the model (i.e., gradient-based methods) or a substantial quantity of data for training a surrogate model. However, under many real-world circumstances, neither of these resources is necessarily available. Our work introduces the concept of the instance-based attack, which is both interpretable and suitable for deployment in a black-box environment. In our approach, a specific binary instance and a malware classifier are utilized as input. By incorporating data augmentation strategies, sufficient data are generated to train a relatively simple and interpretable model. Our methodology involves providing explanations for the detection model, which entails displaying the weights assigned to different components of the specific binary. Through the analysis of these explanations, we discover that the data subsections have a significant impact on the identification of malware. In this study, a novel function preserving transformation algorithm designed specifically for data subsections is introduced. Our approach involves leveraging binary diversification techniques to neutralize the effects of the most heavily-weighted section, thus generating effective adversarial examples. Our algorithm can fool the DNNs in certain cases with a success rate of almost 100%. Instance attack exhibits superior performance compared to the state-of-the-art approach. Notably, our technique can be implemented in a black-box environment and the results can be verified utilizing domain knowledge. The model can help to improve the robustness of malware detectors.