Yaxun Dai, Wenxuan Xie, Xialie Zhuang, Tianyu Yang, Ziyi Liu, Haiqin Yang, Yiying Yang, Yuhang Zhao, Pingfu Chao, Wenhao Jiang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Point-of-Interest (POI) recommendation aims to predict users' future locations based on their historical check-ins. Despite the success of recent deep learning approaches in capturing POI semantics and user behavior, they continue to face the persistent problem of data sparsity and incompleteness. In this paper, we introduce Multi-Objective Adversarial Imitation Recommender (MOAIR), a novel framework that integrates Generative Adversarial Imitation Learning with multi-objective to address this issue. MOAIR effectively captures user behavior patterns and spatial-temporal contextual information via graph-enhanced self-supervised state encoder and overcomes data sparsity by robustly learning from limited data and generating diverse samples. By accommodating diverse user patterns in the training data, the framework also mitigates the typical mode-collapse issue in generative adversarial learning and thus enhances the overall performance. MOAIR employs a multi-objective imitation learning architecture where the imitation learning agent (IL agent) explores the POI space and receives multifaceted reward signals. Utilizing the Paralleled Proximal Policy Optimization (3PO) framework to optimize multi-objectives, the IL agent ensures efficient and stable policy updates. Additionally, to address the issue of high noise in POI recommendation scenarios, we use a novel generative way to define our policy net and incorporate a variational bottleneck for regularization to enhance the stability of adversarial learning. Comprehensive experiments reveal the superior performance for MOAIR compared to other baseline approaches, especially with sparse training data.

Trajectory anomaly detection, as a fundamental operation for moving object pattern discovery, plays an irreplaceable and critical role in spatio-temporal location-based services. Conducting online detection based on the current positions and their contextual semantics can significantly enhance the value of trajectory data. However, existing approaches suffer from two fundamental limitations: 1) treat trajectories as indivisible sequences or apply rigid segmentation strategies, and 2) use of a single detection model that struggles to adapt to concept drift caused by evolving trajectory distributions. Such limitations make it impossible to detect abnormal trajectories in a timely and semantically comprehensive manner. To fill this gap, we propose Stamp, a novel framework for Semantic-aware sub-Trajectory Anomaly detection with a diffusion Multi-model Pool. In particular, Stamp comprises three key innovations: 1. It employs a semantic-driven dynamic segmentation mechanism that identifies natural breakpoints in trajectories based on changes in road semantics, rather than fixed rules. 2. It enhances trajectory representation by embedding road network semantic vectors, capturing both spatial geometry and functional urban characteristics. 3. It employs a pool of diffusion models that dynamically evolves through reliability assessment, similarity measurement, and strategic merging operations, ensuring adaptability to concept drift while leveraging the superior generative capabilities of diffusion models over traditional autoencoders. Experimental results demonstrate that Stamp improves detection efficiency by 35%, AUPR by 5.6%, and F1-score by 2.7% on two large-scale real-world urban trajectory datasets when compared to state-of-the-art methods, demonstrating its effectiveness for real-time anomaly detection in complex urban environments.

Live migration is essential for service continuity of latency-sensitive applications in resource-constrained edge environments. However, current migration approaches for noncontainerized applications suffer from inefficiency, and existing HTTP-based state reporting methods incur substantial overhead while lacking replica pre-initialization capability, resulting in prolonged downtime. To address these limitations, we implement a prototype system that enables near-zero-downtime live migration of object-detection applications. By introducing lightweight, minimally intrusive state management interfaces, our system supports automated application pre-deployment and seamless state transfer. Experimental evaluations on a constrained 50 Mbps edge network demonstrate that object-detection applications experience an average downtime of 132.8 ms, while our solution maintains low resource consumption.

Trajectory similarity aims to identify pairs of similar trajectories, serving as a crucial operation in spatial-temporal data mining. Although several approaches have been proposed, they encounter the following two issues: 1) An overemphasis on spatial similarity in road networks while the rich semantic information embedded in trajectories is not fully exploited; 2) Dependence on Recurrent Neural Network (RNN) architectures would struggle to capture long-term dependencies. To address these limitations, we propose a Dual-branch Attention-based framework with Spatial and Semantic information (DASS) based on self-supervised learning. Specifically, DASS comprises two core components: 1) A trajectory representation module that models spatial-temporal adjacent relationships in the form of graph and converts semantics into numerical embeddings. 2) A backbone encoder with a co-attention module to independently process two features before they are integrated. Extensive experiments on real-world datasets demonstrate that DASS outperforms state-of-the-art methods, establishing itself as a novel paradigm.

Trajectory metric learning, which supports the trajectory similarity search, is one of the most fundamental tasks in spatial-temporal data analysis. However, existing trajectory metric learning methods rely on massive labels of pairwise trajectory distance, and thus cannot be applied to few-shot scenarios frequently occurring in real-world applications. Though performance drops caused by insufficient labels can be alleviated by knowledge distillation, we demonstrate that they cannot be directly applied to few-shot trajectory metric learning due to the domain shift problem. To this end, this paper proposes invariant and relaxed learning enhanced knowledge distillation method TMLKD for few-shot trajectory metric learning, such that domain-invariant representation and rank knowledge can be distilled. Specifically, in the representation learning phase, it first employs an adversarial sub-network to distinguish domain-specific and domain-invariant information, so as to distill transferable representation knowledge from teacher models. To mitigate the few-shot problem in student model training, we further enrich sparse labels of the target domain by utilizing the rank knowledge revealed in teachers' predictions. Particularly, TMLKD employs a list-wise learning-to-rank approach to learn the relaxed trajectory ranking orders instead of focusing on all the samples inefficiently. Finally, to guide accurate distillation, we adaptively assign reliability of teacher prediction by utilizing the ground-truth labels, to avoid misleading the student model with low-quality teacher predictions. Extensive experiments on three real-world datasets demonstrate the superiority of our model.

Trajectory similarity is a fundamental operation in spatial-temporal data mining with wide-ranging applications. However, trajectories inherently exhibit diversity due to varied sampling and distribution of trajectory points, influenced by different motion patterns, sampling methods, and route constraints. This diversity leads to varying results in trajectory similarity measures, including DTW, LCSS, ED, Hausdorff, EDR, etc. In this paper, we argue for a comprehensive consideration of various distance metrics to enhance accuracy. To address this, this paper proposes a Multi-source Deep Trajectory Representation Learning method for accurate and efficient similarity queries. In particular, MDTRL comprises two key modules: (1) A novel trajectory representation module that incorporates an attention-based embedding mechanism and a deep metric learning network aggregating multiple measures. (2) A continuous metric learning strategy that adaptively updates similarity, thereby enhancing the accuracy of similarity queries. We employ the locality sensitive hashing index to further improve the similarity query. Extensive experiments conducted on real trajectory datasets reveal that MDTRL has state-of-the-art solutions, in terms of both effectiveness and efficiency across multi-source trajectories. It achieves 5x-15x speedup and 10%-15% accuracy improvement over Euclidean, Hausdorff, DTW, and Discrete Fréchet measures.

Distantly supervised named entity recognition (DS-NER) has emerged as a cheap and convenient alternative to traditional human annotation methods, enabling the automatic generation of training data by aligning text with external resources. Despite the many efforts in noise measurement methods, few works focus on the latent noise distribution between different distant annotation methods. In this work, we explore the effectiveness and robustness of DS-NER by two aspects: (1) distant annotation techniques, which encompasses both traditional rule-based methods and the innovative large language model supervision approach, and (2) noise assessment, for which we introduce a novel framework. This framework addresses the challenges by distinctly categorizing them into theunlabeled-entity problem (UEP)and thenoisy-entity problem (NEP), subsequently providing specialized solutions for each. Our proposed method achieves significant improvements on eight real-world distant supervision datasets originating from three different data sources and involving four distinct annotation techniques, confirming its superiority over current state-of-the-art methods.

With the popularization of mobile applications and the timely acquisition of fresh data, real-time clustering and its evolution analysis have become the primary operations for data processing and knowledge discovery. Such continuous queries on massive objects are computation-intensive tasks in dynamic scenarios. However, existing clustering techniques are incompetent to achieve decent performance when computation-intensive operations frequently occur in streaming scenarios, which is caused by two challenges: (i) uncertainty of the clustering frequency; (ii) unpredictable distribution evolution. Hence, it is critical to find a lightweight model that can cluster the high-speed dynamic instances while exploiting the evolution amid different clustering results. This paper focuses on the problem of real-time clustering on streaming data in computation-intensive and high-dynamics tasks, through a framework Ocean, consisting of the Online clustering algorithm and evolution analysis. Particularly, the framework conceives a flexible composite window to augment the knowledge mining, achieving a proper real-time response in various scenarios. The evolution analysis supports full life-cycle detection, improving the adaptability to dynamic concept drifts and multiple patterns. Inspired by the grid partition strategy, this framework adopts grid feature vectors to capture the significant changes in streaming data. Furthermore, we propose an optimization that removes sparse grids timely and performs the online clustering adaptively for space and time efficiency. It is proven to be effective both theoretically and experimentally. This strategy enables real-time clustering for dynamic streaming data without degrading the clustering quality or increasing the computation cost. Experiments on real datasets and synthetic datasets verify the accuracy and effectiveness of Ocean compared to the state-of-the-art approaches, as well as the superior ability to perform clustering in a real-time manner.

Spatial trajectory prediction is a fundamental problem for diverse location-based applications. However, existing methods fall short in learning and generalization, and cannot sufficiently capture users’ spatiotemporal preferences, especially for cold-start users. Moreover, these methods do not explicitly consider the diversity of moving patterns among users and trajectories, i.e., the learning difficulty of different user and trajectory samples, thus hindering the improvement of prediction accuracy. To solve these problems, we propose a novel Curriculum and Contrastive Learning Enhanced Meta-Learner (CCML) that transfers knowledge from users with rich data to cold-start users. Specifically, a Contrastive-based Trajectory Predictor (CTP) is designed as the base model, which utilizes contrastive learning technique on both user-level and trajectory-level, aiming to facilitate a more profound understanding and differentiation of the varied travel behaviors and preferences exhibited by individuals. Meanwhile, CCML also incorporates the curriculum learning and the hard sample mining strategies. It simultaneously considers the learning difficulty of both user and trajectory samples, and presents the learning tasks by an easy-to-hard curriculum. By learning more challenging combinations of user and trajectory samples in each meta-learning iteration, the meta-learner can converge to a better status. Extensive experiments on two real-world datasets demonstrate the superiority of our models.

Map matching is a problem to align recorded location data to a digital map. It has been well studied to map GPS data collected from vehicles to paths in a road network. The problem of Cellular Trajectory Map-Matching (CTMM) is a new problem that deals with trajectories of cellular-based positioning data. It has a wide range of applications, for example, for telecommunication companies to understand and predict traffic information based on telecom tokens obtained from vehicles. CTMM is a significantly more challenging task that faces much lower data precision and higher positioning errors. While Hidden Markov Model (HMM) based methods can achieve satisfactory results for GPS-based map matching, we show that they cannot be directly applied to the CTMM problem. In this paper, we aim at reducing the impact of positioning errors by incorporating knowledge obtained by neural networks into learned probabilities. A multi-relational graph learning method is developed to generate meaningful embedding, with multi-relational useful information fully preserved in a shared space. An attentive neural network is then designed as the learner for observation probability, incorporating the knowledge of the dynamic correlation between roads and cell towers under varying trajectory contexts. A transition probability learner is used to capture implicit deep features for enhanced transition probability modeling. Finally, the learned observation and transition probabilities are seamlessly integrated into HMM to guide more accurate path-finding. Extensive experiments on two large-scale cellular datasets reveal that our approach achieves high accuracy and robustness on CTMM.

With the popularity of modern mobile devices and GPS technology, big web stream data with location are continuously generated and collected. The sequential positions form a trajectory, and the clustering analysis on trajectories is beneficial to a wide range of applications, e.g., route recommendation. In the past decades, extensive efforts have been made to improve the efficiency of static trajectory clustering. However, trajectory stream data is received incrementally, and the continuous trajectory clustering inevitably faces the following two problems: (1) physical structure design for trajectory representation leads to severe space overhead, and (2) dynamic maintenance of trajectory semantics and its retrieval structure brings intensive computation. To overcome the above problems, an adaptive continuous trajectory clustering framework (ACTOR) is proposed in this paper. Overall, it covers three key components: (1) Simplifier represents trajectory with a well-designed PT structure. (2) Partitioner utilizes a hexagonal-based indexing strategy to enhance the local computational efficiency. (3) Executor accommodates an adaptive selection of P-clustering and R-clustering approaches according to the ROC (rate of change) matrix. Empirical studies on real-world data validate the usefulness of our proposal and prove the huge advantage of our approach over available solutions in the literature.

Trajectory data has become ubiquitous nowadays, which can benefit various real-world applications such as traffic management and location-based services. However, trajectories may disclose highly sensitive information of an individual including mobility patterns, personal profiles and gazetteers, social relationships, etc, making it indispensable to consider privacy protection when releasing trajectory data. Ensuring privacy on trajectories demands more than hiding single locations, since trajectories are intrinsically sparse and high-dimensional, and require to protect multi-scale correlations. To this end, extensive research has been conducted to design effective techniques for privacy-preserving trajectory data publishing. Furthermore, protecting privacy requires carefully balance two metrics: privacy and utility. In other words, it needs to protect as much privacy as possible and meanwhile guarantee the usefulness of the released trajectories for data analysis. In this survey, we provide a comprehensive study and a systematic summarization of existing protection models, privacy and utility metrics for trajectories developed in the literature. We also conduct extensive experiments on two real-life public trajectory datasets to evaluate the performance of several representative privacy protection models, demonstrate the trade-off between privacy and utility, and guide the choice of the right privacy model for trajectory publishing given certain privacy and utility desiderata.

Map inference algorithm aims to construct a digital map from other data sources automatically. Due to the labour intensity of traditional map creation and the frequent road change nowadays, map inference is deemed to be a promising solution to automatic map construction and update. However, existing map inference from GPS trajectories suffers from low GPS data quality, which makes the quality of the constructed map unsatisfactory. In this paper, we study the existing map inference algorithms using GPS trajectories. Different from previous surveys, we (1) include the most recent solutions and propose a new categorisation of method; (2) study how different types of GPS errors affect the quality of inference results; (3) evaluate the existing map inference quality measures regarding their ability to identify map quality issues. To achieve these goals, we conduct a comprehensive experimental study on several representative algorithms using both real-world datasets and synthetic datasets, which are generated from our proposed synthetic trajectory generator and artificial map generator. Overall, our study provides insightful observations regarding (1) which inference method performs better in each working scenario, (2) the general data quality requirements for map inference, (3) the direction of future works for quantitative map quality measures.

The Constrained Shortest Path (CSP) problem aims to find the shortest path between two nodes in a road network subject to a given constraint on another attribute. It is typically processed as a skyline path problem on the two attributes, resulting in very high computational cost which can be prohibitive for large road networks. The main bottleneck is to deal with a large amount of partial skyline paths, which further makes the existing index-based methods incapable to obtain the complete exact skyline paths. In this paper, we propose a novel skyline path concatenation approach to avoid the expensive skyline path search, which is then used to efficiently construct a 2-hop labeling index for the CSP queries. Specifically, a rectangle-based technique is designed to prune the concatenation space from multiple hops, and a constraint pruning method is used to further speed up the CSP query processing. To further scale up to larger networks, we propose a novel forest hop labeling that constructs labels from different partitions in parallel. Our approach is the first method that can achieve both accuracy and efficiency for CSP query answering. Extensive experiments on real-life road networks demonstrate that our method outperforms the state-of-the-art CSP solutions by several orders of magnitude.

Shortest path finding is the building block of various applications in road networks and the index-based algorithms, especially hub labeling, can boost the query performance dramatically. However, the traffic condition keeps changing in real life, making the pre-computed index unable to answer the query correctly. In this work, we adopt the state-of-the-art tree decomposition-based hub labeling as the underlying index, and design efficient algorithms to incrementally maintain the index. Specifically, we first analyze the structural stability of the index in dynamic road networks which enables us to concentrate on label value maintenance. We then introduce the minimum weight property and minimum distance property to guarantee the index correctness without graph traversal. Moreover, we propose the star-centric paradigm for tracing index change and design various pruning techniques to further accelerate the index maintenance. Finally, we extend our algorithms to batch mode for shared computation, extend to structural maintenance for full types of update, and generalize to all kinds of TDHL. Our experimental results validate the superiority of our proposals over existing solutions on both index maintenance and query processing.

Next Point-of-Interest (POI) recommendation is of great value for location-based services. Existing solutions mainly rely on extensive observed data and are brittle to users with few interactions. Unfortunately, the problem of few-shot next POI recommendation has not been well studied yet. In this paper, we propose a novel meta-optimized model MFNP, which can rapidly adapt to users with few check-in records. Towards the cold-start problem, it seamlessly integrates carefully designed user-specific and region-specific tasks in meta-learning, such that region-aware user preferences can be captured via a rational fusion of region-independent personal preferences and region-dependent crowd preferences. In modelling region-dependent crowd preferences, a cluster-based adaptive network is adopted to capture shared preferences from similar users for knowledge transfer. Experimental results on two real-world datasets show that our model outperforms the state-of-the-art methods on next POI recommendation for cold-start users.

Route reconstruction targets at recovering the actual routes of objects moving on an underlying road network from their times-tamped position measurements. This fundamental pre-processing step to many location-based applications has been extensively studied for GPS data, which are object-centric and relatively densely sampled data. In this paper, we investigate the problem of route reconstruction using data collected from road-side Bluetooth scanners. In many cities, Bluetooth scanners are installed in road networks for monitoring the movement of Bluetooth-enabled devices. To address new challenges caused by such reader-centric Bluetooth data including spatial and temporal distortion, a new route reconstruction framework is proposed to transform Bluetooth readings through a family of distortion suppression strategies such that the transformed data can work well with the Hidden Markov model (HMM) map-matching approach. Extensive experiments are conducted to evaluate different transformation strategies with real-world datasets. The experimental results show that when the algorithm uses the baseline or the proposed transformation strategies, the map matching F1 score can be increased by up to 10% depending on the severity of distortion.