Interactive visual query systems are essential for the exploration and analysis of geospatial data. However, developing such systems has become increasingly challenging in recent years due to the conflict between the unprecedented volume of data and the need for instantaneous feedback. To address this challenge, we present HiVQ, a High-performance Visual Query system for real-time interactive visual query of geospatial big data. HiVQ adopts an innovative “Query as Visualization” paradigm, transforming user interactions into pixel value queries which can be processed efficiently with specialized indices and optimization strategies. Unlike conventional solutions that query and visualize geospatial objects sequentially, HiVQ effectively omits most geospatial objects and unnecessary computations that do not affect the final visualization, ensuring minimal sensitivity to data volume. Experimental results show that HiVQ accelerates visual queries by at least seven times compared to SOTA methods. This demonstration enables users to interactively explore and analyze spatial data with billions of nodes at any scale, receiving responses in milliseconds as they dynamically adjust analysis parameters, query conditions, or map styling. The demonstration video is available at https://gitee.com/kyrie-Bang/HiVQ-Demo.

In the rapidly evolving global shipping industry, accurate vessel trajectory prediction is essential for effective maritime traffic management. However, the inherent uncertainties in vessel voyages pose significant challenges for existing methods in achieving precise long-term predictions. Inspired by the success of large language models (LLMs) in natural language processing, we propose SEMINT—a novel framework that integrates SEMantic cognition and INTent context-awareness for long-term vessel trajectory prediction. SEMINT leverages LLMs to semantically interpret vessel navigation behaviors and infer multiple voyage intents. The final intent inference is derived from the consensus among multiple reasoning chains, which are then combined with historical preferences to form an intent context. Guided by this context, a Transformer-based trajectory prediction model captures the dependencies between critical waypoints and voyage intent, thereby achieving stable long-term numerical predictions. SEMINT combines the advantages of LLMs and task-specific deep learning models, transforming voyage uncertainties into high-level intent diversity and simplifying the learning process for long-term movement patterns. Experiments on real-world datasets show that SEMINT outperforms previous methods, reducing average and final displacement errors by 28.47% and 31.59%, respectively.

Knowledge graph is becoming an indispensable resource that offers structured information for numerous AI applications. However, the knowledge graph often suffers from its incompleteness. Building a complete, high-quality knowledge graph is time-consuming and requires significant human annotation efforts. In this paper, we study the Knowledge Graph Population task, which aims at extending the scale of structured knowledge, with a special focus on reducing data preparation and annotation efforts. Previous works mainly based on discriminative methods build classifiers and verify candidate triplets that are extracted from texts, which heavily rely on the quality of data collection and co-occurrance of entities in the text. However, such methods fail to generalize on entity pairs that are not highly co-occurred, and fail to discover entity pairs that are not co-occurred at all in the given text corpus. We introduce a generative perspective to approach this task and define each relationship by learning the data distribution that embodies the core common properties for relational reasoning. A generative modelKGGenis proposed, which samples from the learned data distribution for each relation and can generate triplets regardless of entity pair co-occurrence in the text corpus. To further improve the generation quality while alleviate human annotation efforts, adversarial learning is adopted to not only encourage generating high quality triplets, but also give model the ability to automatically assess the generation quality. Quantitative and qualitative experimental results conducted on two real-world generic knowledge graphs show that the proposed modelKGGengenerates novel and meaningful triplets with improved efficiency and less human annotation comparing with the state-of-the-art approaches.

Query feeds recommendation is a new recommended paradigm in mobile search applications, where a stream of queries need to be recommended to improve user engagement. It requires a great quantity of attractive queries for recommendation. A conventional solution is to retrieve queries from a collection of past queries recorded in user search logs. However, these queries usually have poor readability and limited coverage of article content, and are thus not suitable for the query feeds recommendation scenario. Furthermore, to deploy the generated queries for recommendation, human validation, which is costly in practice, is required to filter unsuitable queries. In this paper, we propose TitIE, a query mining system to generate valuable queries using the titles of documents. We employ both an extractive text generator and an abstractive text generator to generate queries from titles. To improve the acceptance rate during human validation, we further propose a model-based scoring strategy to pre-select the queries that are more likely to be accepted during human validation. Finally, we propose a novel dual learning approach to jointly learn the generation model and the selection model by making full use of the unlabeled corpora under a semi-supervised scheme, thereby simultaneously improving the performance of both models. Results from both offline and online evaluations demonstrate the superiority of our approach.

We present DiSA, a Display-driven Spatial Analysis framework for interactive analysis of large-scale geographical vector data. DiSA calculates visualization of analysis results directly using a parallel per-pixel approach with efficient fine-grained spatial indexes. Compared with conventional object-based methods, DiSA can greatly reduce the computational complexity (from O(n) to O(log(n)) in some cases), making it less sensitive to data volumes. Experimental results verify that DiSA can provide analysis of billion-scale spatial objects in milliseconds. We demonstrate DiSA with various application scenarios including raw data exploration, spatial buffer and overlay analysis, and global cellular signal strength analysis. Users can explore 10 millions of spatial objects, adjust algorithm parameters, and always see the results in real-time on a personal computer.

Aiming at the existing problem of single buffer replacement policy and the inability of predicting future SQLs execution in current database buffer replacement algorithm, proposed a new database buffer replacement algorithm---CG-ABR (call graph optimizing adaptive buffer replacement algorithm). CG-ABR could adapt to current SQL's access behavior automatically by getting it through query optimizer, and this algorithm optimized the management method of buffer access information, and generated the call graph according to the SQL's execution order of application program's logic, based on this call graph, the future SQLs execution can be predicted, and the pages might not be accessed hereafter, would be replaced, which could optimize the buffer space. The experiments and analysis showed that CG-ABR algorithm is effective in buffer replacement, with high hit ratio to real time access types and enormous value in practicing applications.

This paper introduces an efficient non-blocking spatial join (NSJ, for short) algorithm to deal with spatial objects from remote sources via underlying network. The objectives of NSJ are: (1) start reporting the first output join results as soon as possible, and (2) minimize the cost for output the remaining results. As some other previous non-blocking join algorithms, NSJ includes two stages: memory-join stage and disk-join stage The memory-join stage employs a join process as along as receiving spatial objects, and the disk-join stage is responsible for the uncompleted join process during the memory-join stage after all spatial objects are received completely. We propose a dynamic concurrent flush policy(DCFP) based on resident degree to process memory overflow, which makes join process in memory-join stage more efficiently. We also develop an optimal data access schedule algorithm based on BEA (Bond Energy Algorithm) to reduce redundant I/O and CPU cost in disk-join stage. Extensive experiments prove that our technique delivers result significantly more efficient than the previous methods.

Efficient path computation is essential for applications such as intelligent transportation systems (ITS) and network routing. In ITS navigation systems, many path requests can be submitted over the same, typically huge, transportation network within a small time window. While path precomputation (path view) would provide an efficient path query response, it raises three problems which must be addressed: 1) precomputed paths exceed the current computer main memory capacity for large networks; 2) disk-based solutions are too inefficient to meet the stringent requirements of these target applications; and 3) path views become too costly to update for large graphs (resulting in out-of-date query results). We propose a hierarchical encoded path view (HEPV) model that addresses all three problems. By hierarchically encoding partial paths, HEPV reduces the view encoding time, updating time and storage requirements beyond previously known path precomputation techniques, while significantly minimizing path retrieval time. We prove that paths retrieved over HEPV are optimal. We present complete solutions for all phases of the HEPV approach, including graph partitioning, hierarchy generation, path view encoding and updating, and path retrieval. In this paper, we also present an in-depth experimental evaluation of HEPV based on both synthetic and real GIS networks. Our results confirm that HEPV offers advantages over alternative path finding approaches in terms of performance and space efficiency.

Investigates optimization strategies for processing path queries with embedded spatial constraints, such as avoiding areas with certain characteristics. To resolve complex spatial constraints during path finding, we consider two decisions: (1) the spatial relation operations (e.g. intersection) between areas and links can be pre-processed or intermixed with path-finding, and (2) areas satisfying the query constraint can be pre-filtered or dynamically selected during path-finding. Based on these two decisions, we propose and implement the resulting four integrated query processing strategies, utilizing state-of-the-art technologies such as spatial joins for intersect computation, R-tree access structure for spatial overlap searching, and spatial clustering for efficient path searching. In this paper, we also report an experimental evaluation to show which strategies perform best in different scenarios.

The development of a cost model for predicting the performance of spatial joins has been identified in the literature as an important and difficult problem. The authors present the first cost model that can predict the performance of spatial joins using R-trees. Based on two existing R-trees (join targets), the model first estimates the number of expected I/Os for the join process by assuming a zero buffer size. The method for this estimation extends the cost model for R-tree window queries (developed by Kamel and Faloutsos (1993) and by Pagel et al. (1993)) to also handle spatial joins (which are more complex). In the context of spatial join processing, this number of zero-buffer expected I/Os is not practical for performance prediction in a buffered environment. To model the buffer impact, they use an (exponential) distribution function to measure the probability that a bufferless I/O would cause a page fault in a buffered environment. Based on this probability and the zero-buffer expected I/O cost, the estimated number of I/Os for an R-tree join can then be computed. The comparisons between the predictions from the cost model and the actual results from the experiments based on real GIS maps show that the average relative error ratio is about 10% with a maximum of about 20% for a wide range of buffer sizes. Therefore, our model is a useful tool for the query optimization of spatial join queries.

In this paper, we present an experimental evaluation of graph clustering strategies in terms of their effectiveness in optimizing I/O for path query processing in digital map databases.Clustering optimization is attractive because it does not incurs any run-time cost, and is complimentary to many of the existing techniques in path query optimization.We first propose a novel graph clustering technique, called Spatial Partition Clustering (SPC), that creates balanced partitions of links based on the spatial proximity of their origin nodes.We then select three alternative clustering techniques from the literature, namely two-way partitioning, approximately topological clustering, and random clustering, to compare their performance in path query processing with SPC.Experimental evahration indicatesthat our SPC performs the best for the high-locality graphs (such as GIS maps), whereas the two-way partitioning approach performs the best for no-locality random graphs.

Efficient path query processing is a key requirement for advanced database applications including GIS (Geographic Information Systems) and ITS (Intelligent Transportation Systems).We study the problem in the context of automobile navigation systems where a large number of path requests can be submitted over the tmrtsportation network within a short period of time.To guarantee efficient responsefor path queries, we employ a path view materiatization strategy for precomputing the best paths.We tackle the following three issues: (1) memory-resident solutions quickly exceed cr.nrentcomputer storage capacity for networks of thousands of nodes, (2) diskbased solutions have been found inefficient to meet the stringent performance requirements, and (3) path views become too costly to update for large graphs.We propose the HEPV (Hierarchical Encoded Path View) approach that addressesthese problems while guaranteeing the optimality of path retrieval.Our experimental fesuks reveaf that HEPV is more efficient than previously known path finding approaches.