Location-Based Social Network (LBSN) check-in trajectory data are important for many practical applications like POI recommendation, advertising, and pandemic intervention. However, the high collection costs and ever-increasing privacy concerns prevent us from accessing large-scale LBSN trajectory data. The recent advances in synthetic data generation provide us with a new opportunity to achieve this, which utilizes generative AI to generate synthetic data that preserves the characteristics of real data while ensuring privacy protection. However, generating synthetic LBSN check-in trajectories remains challenging due to their spatially discrete, temporally irregular nature and the complex spatio-temporal patterns caused by sparse activities and uncertain human mobility. To address this challenge, we propose GeoGen, a two-stage coarse-to-fine framework for large-scale LBSN check-in trajectory generation. In the first stage, we reconstruct spatially continuous, temporally regular latent movement sequences from the original LBSN check-in trajectories and then design a Sparsity-aware Spatio-temporal Diffusion model (S^2TDiff) with an efficient denosing network to learn their underlying behavioral patterns. In the second stage, we design Coarse2FineNet, a Transformer-based Seq2Seq architecture equipped with a dynamic context fusion mechanism in the encoder and a multi-task hybrid-head decoder, which generates fine-grained LBSN trajectories based on coarse-grained latent movement sequences by modeling semantic relevance and behavioral uncertainty. Extensive experiments on four real-world datasets show that GeoGen excels state-of-the-art models for both fidelity and utility evaluation, e.g., it increases over 69% and 55% in distance and radius metrics on the FS-TKY dataset.

Urban mobility data has significant connections with economic growth and plays an essential role in various smart-city applications. Due to privacy concerns and substantial data collection costs, fine-grained human mobility trajectories are challenging to make publicly available on a large scale. A promising solution to address this issue is trajectory synthesizing, which generates synthetic trajectories that preserve aggregate spatiotemporal distributions. However, existing works often neglect the road network structural constraints or rely on instance-level external supervision, thus limiting their scalability in generating fine-grained and high-fidelity trajectories. In this paper, we propose Cardiff, a coarse-to-fine Cascaded hybrid diffusion-based framework for fine-grained and structure-plausible trajectory generation. By leveraging the hierarchical nature of urban mobility, Cardiff decomposes the generation process into two cascaded levels, i.e., discrete road segment-level and continuous fine-grained GPS-level: (i) At the segment level, to reduce computational costs and redundancy in raw trajectories, we first encode the discrete road segments into low-dimensional latent embeddings and design a diffusion transformer-based latent denoising network for segment-level synthesis. (ii) Taking the first stage of generation as conditions, we then design a fine-grained GPS-level conditional denoising network with a noise augmentation mechanism to achieve road-network-constrained and fine-grained generation. The cascaded progressive generation yields high-fidelity fine-grained trajectories while adhering to road geometry and topology constraints. Experimental results on three large real-world trajectory datasets demonstrate that our method outperforms state-of-the-art baselines in various metrics. The code is available at~ https://github.com/urban-mobility-generation/Cardiff.

Points of Interest (POIs) are crucial for urban planning and location-based services. However, the high dynamics and expensive updating costs of POIs pose a key roadblock for their urban applications. This is especially true for developing countries, where active economic activities lead to frequent POI updates (e.g., merchants closing down and new ones opening). Therefore, POI updating, i.e., detecting new POIs and different names of the same POIs (alias) to update the POI database, has become an urgent but challenging problem to address. In this paper, we attempt to answer the research question of how to detect and update large-scale POIs via a low-cost approach. To do so, we propose a novel mobile sensing framework called UrbanPOI, which consists of three key modules: (i) a hierarchical POI candidate generation module based on the POINet model that detects POIs from shipping addresses; (ii) a new POI detection module based on the Siamese Attention Network that models multi-modal data and crowd intelligence; and (iii) a POI profiling model to infer the semantic category and the spatial location of updated POIs. Extensive results on real-world logistics delivery datasets show that our model outperforms state-of-theart models in Beijing City by 26.2% in precision and 10.7

Last-mile delivery from delivery stations to customers is now mainly finished by dedicated couriers. In practice, each courier generally collects orders destined for one delivery area at the station and delivers them. However, the long distance between the station and the delivery area, due to practical reasons such as the expensive station rental fee in the downtown area, increases the travel time and decreases delivery efficiency. We solve the problem with relay transportation, where one or multiple relay couriers collect orders at the station and send them to delivery couriers, and delivery couriers focus on order delivery in the corresponding areas. We design a mobility-aware single relay courier scheduling system calledREDEto minimize the average relay order delivery time (ARODT) for last-mile delivery. First, a heterogeneous task aware route prediction algorithm is proposed to characterize the delivery courier's mobility. After that, a distance-aware greedy algorithm and an ARODT-constrained exchange algorithm are designed to generate the relay route. A hybrid last-mile delivery systemREDE+is then designed to further reduce ARODT ofREDE. We also proposeREDE-COto optimize multiple relay courier cooperative scheduling for order relay. Extensive evaluation shows the effectiveness ofREDE,REDE+, andREDE-COin reducing ARODT.

Singing voice synthesis has made remarkable progress in generating natural and high-quality voices. However, existing methods rarely provide precise control over vocal techniques such as intensity, mixed voice, falsetto, bubble, and breathy tones, thus limiting the expressive potential of synthetic voices. We introduce TechSinger, an advanced system for controllable singing voice synthesis that supports five languages and seven vocal techniques. TechSinger leverages a flow-matching-based generative model to produce singing voices with enhanced expressive control over various techniques. To enhance the diversity of training data, we develop a technique detection model that automatically annotates datasets with phoneme-level technique labels. Additionally, our prompt-based technique prediction model enables users to specify desired vocal attributes through natural language, offering fine-grained control over the synthesized singing. Experimental results demonstrate that TechSinger significantly enhances the expressiveness and realism of synthetic singing voices, outperforming existing methods in terms of audio quality and technique-specific control.

The rapid growth of e-commerce has made last-mile delivery a critical service in daily life. Despite regulations mandating doorstep delivery, the pressure of penalties for delays can lead to fraudulent delivery behaviors, where couriers may report package receipt without actually deliver the package to assigned locations. Existing studies on fraud behavior detection focus on exploring user (courier) behaviors for fraud behavior detection. However, due to the inaccuracy of GPS positioning and the variability of user behavior patterns caused by dynamic environmental factors, relying solely on behavior data remains insufficient for detecting fraudulent deliveries. In this paper, we present a Multimodal Fraudulent Delivery Detection framework (MFDD), which integrates heterogeneous data from multiple agents (courier-side and user-side)-including couriers' physical behavior, digital behavior, and conversations containing customer feedback-for detecting fraudulent deliveries in the last-mile delivery. We employ attention mechanisms to extract features from each modality and use cross-modal fusion to capture complex and varied relationships between multimodal data. To further mitigate modality imbalance during training, we introduce a dynamic gradient-modulation strategy that balances learning across all modalities. We implement and evaluate MFDD on real-world, human-annotated data, achieving a 9.6% improvement in precision and a 5.8% increase in accuracy over the state-of-the-art methods. We also deploy the model in the production environment of JD Logistics, and results show that compared to existing methods, MFDD improves accuracy by 15.3%, reducing estimated annual costs by over 18.5 million CNY.

In last-mile logistics, couriers are typically assigned fixed zones to perform door-to-door deliveries. In practice, packages in some delivery zones might not be fulfilled on time due to couriers taking irregular leave for sickness or higher-priority task assignments, e.g., services for VIPs and regulatory training. Beyond the costly real-world practice, i.e., hiring temporary workers, analysis of historical data reveals that a daily delivery task rescheduling among on-duty couriers can be a cost-effective and efficient alternative. It involves individual workload assessments and delivery task assignments, both of which existing methods can not address adequately: (i) Existing courier workload assessment methods are not tailored for downstream optimization tasks, leading to poor performance. (ii) Efficiency-oriented task assignment methods may lead to unfair workload among the couriers. To address the above two limitations, in this paper, we propose D3-TR, a data-driven method for task reassignment among present couriers. Firstly, we design a consistency-guided predictor that can quickly and precisely predict the workload of couriers. Secondly, based on this predictor, we design a workload-aware genetic algorithm to solve the optimal task allocation problem. Experimental results underscore the superiority of our method over several baselines. Furthermore, real-world deployment on millions of orders demonstrates the effectiveness of our solution, yielding an average of 3.9% improvement in the on-time delivery rate.

Non-emergency municipal services, e.g., city 311 systems, have been widely implemented across cities in Canada and the United States to enhance residents' quality of life. These systems enable residents to report issues, e.g., noise complaints, missed garbage collection, and potholes, via phone calls, mobile applications, or webpages. However, residents are often given limited information about when their service requests will be addressed, which can reduce transparency, lower resident satisfaction, and increase the number of follow-up inquiries. Predicting the service time for municipal service requests is challenging due to several complex factors: (i) dynamic spatial-temporal correlations, (ii) underlying interactions among heterogeneous service request types, and (iii) high variation in service duration even within the same request category. In this work, we propose MuST2-Learn: a Multi-view Spatial-Temporal-Type Learning framework designed to address the aforementioned challenges by jointly modeling spatial, temporal, and service type dimensions. In detail, it incorporates an inter-type encoder to capture relationships among heterogeneous service request types and an intra-type variation encoder to model service time variation within homogeneous types. In addition, a spatiotemporal encoder is integrated to capture spatial and temporal correlations in each request type. The proposed framework is evaluated with extensive experiments using two real-world datasets. The results show that MuST2-Learn reduces mean absolute error by at least 32.5%, which outperforms state-of-the-art methods.

3D spatial sound field synthesis takes the head-mounted audio signals and body poses as input and renders a 3D sound field around the center body, in which spatial audio can be inferred at any arbitrary position. To achieve this, a multi-modal system is required to spatialize input audio signals with the guidance of pose streams. Then, a numerical method is used to render the sound field with hundreds of spatial audio. However, there exist several challenges including hybrid audio input signals, cross-modal matching, and few data resources. To address them, we propose NAT3DSound, a unit-based framework for high-quality 3D spatial sound synthesis, which consists of 1) a general tokenization scheme for both dense and impulsive input audio signals, 2) a non-autoregressive transformer with multi-scale modality fusion for efficient cross-modal alignment and 3) a parallel sampling strategy for fast prediction. Furthermore, we investigate the feasibility of acoustic pre-training for low-resource learning in data-scarce scenarios. Extensive experiments and ablation studies demonstrate the effectiveness of NAT3DSound in terms of spatialization quality and generalization ability. Audio samples are available at http://NAT3DSound.github.io

A long-standing challenge for pushing sensor-based human activity recognition (HAR) to industrial usage is the distribution shift between training data and testing data: significant variations in data distribution lead to a notable decline in performance. Recently, Large Language Models (LLMs) have demonstrated exceptional generalization capability, which provides a new opportunity to mitigate the distribution shift problem of HAR. However, since LLMs are inherently designed and trained on textual data, their potential to enhance generalization in HAR applications remains an open question. In this paper, we introduce LLM4HAR, a novel LLM-based model to improve cross-domain HAR. LLM4HAR consists of three main modules: (i) the Sensor Data Adaptation module, which aligns IMU signals with LLMs via sensor embedding(ii) the Sensor Knowledge Learning module, which injects sensor knowledge into LLMs for activity recognition, and (iii) the Efficiency Enhancement module, which employs a partial training strategy and reduces the model size by more than 10 times. Extensive evaluations show that LLM4HAR outperforms the existing methods by 13.82% in average F1 score, demonstrating the feasibility and effectiveness of transferring knowledge from pretrained LLMs to enhance HAR. Further, LLM4HAR has been adopted by JD Logistics to support downstream applications such as Courier Welfare Improvement and Map Data Generation.

An emerging type of fraud involves malicious senders exploiting the blind shipment and cash-on-delivery (COD) mechanisms by dispatching large volumes of unsolicited, low-cost parcels. If unsuspecting receivers accept these parcels, they pay for both shipping and goods; otherwise, logistics providers bear the round-trip shipping costs. Existing detection techniques, which rely on extensive labeled cases, struggle with this emerging fraud because receivers' unawareness and low transaction values discourage complaints, resulting in few confirmed cases. Therefore, we propose leveraging receivers' complaints, though not initially collected for fraud detection, to uncover subtle indicators of fraud patterns, while addressing three challenges: (C1) noise-rich dialogues(C2) data privacy concerns, and (C3) ever-evolving fraud patterns. To address them, we design BLOFF, a Blind shipment detection Framework for LO gistics Fraud powered by large language models (LLMs). Specifically, BLOFF includes three components: i) Sensitivity Anonymization to protect sensitive user information; ii) Dialogue Profile Distillation to transform informal dialogues into structured representation, addressing C1, and distill knowledge from a teacher LLM (GPT-4o) to a lightweight student LLM (ChatGLM4-9B), addressing C2; ii) Multi-faceted Context Augmentation to enhance the interpretation of fraud signatures and adaptation of evolving patterns, addressing C3. We evaluate BLOFF on about 56,000 complaints records collected from JD Logistics between January and November 2024. Results show that BLOFF outperforms state-of-the-art methods, achieving a 10.19% improvement in precision. Furthermore, during its real-world deployment in December 2024, BLOFF identified over 90 fraudulent parcels with a 91.4% precision.

Textual description of a physical location, commonly known as an address, plays an important role in location-based services(LBS) such as on-demand delivery and navigation. However, the prevalence of abnormal addresses, those containing inaccuracies that fail to pinpoint a location, have led to significant costs. Address rewriting has emerged as a solution to rectify these abnormal addresses. Despite the critical need, existing address rewriting methods are limited, typically tailored to correct specific error types, or frequently require retraining to process new address data effectively. In this study, we introduce AddrLLM, an innovative framework for address rewriting that is built upon a retrieval augmented large language model. AddrLLM overcomes aforementioned limitations through a meticulously designed Supervised Fine-Tuning module, an Address-centric Retrieval Augmented Generation module and a Bias-free Objective Alignment module. To the best of our knowledge, this study pioneers the application of LLM-based address rewriting approach to solve the issue of abnormal addresses. Through comprehensive offline testing with real-world data on a national scale and subsequent online deployment, AddrLLM has demonstrated superior performance in integration with existing logistics system. It has significantly decreased the rate of parcel re-routing by approximately 43%, underscoring its exceptional efficacy in real-world applications.

Human behavior sensing has been receiving growing attention from both academia and industry in recent years. We report - to the best of our knowledge - the first AI-driven nationwide human behavior sensing system called SMILE in urban last-mile delivery. SMILE detects real-time human behaviors with self-supervised sensor data pretraining and uploads detection results to cloud servers with mobile networks. During its full deployment phase at JD Logistics, SMILE is deployed on over 500,000 mobile devices carried by over 300,000 delivery couriers who travel more than 10 million KM every day in 366 Chinese cities. SMILE serves the delivery of 7 billion E-commerce orders every year for more than 500 million customers. SMILE detects walking, upstairs, downstairs, still, and driving behaviors. SMILE has been fully deployed at JD Logistics to support two real-world applications that benefit both the delivery couriers and the logistics platform: (1) workload measurement to improve couriers' welfare; (2) large-scale delivery map data generation to improve the logistics delivery efficiency.

Wireless sensing technologies have shown significant promise in various applications, but their spatial coverage is confined to the vicinity of the transmitters, limiting their applicability in broader environments. In this paper, we introduce an innovative wireless sensing approach based on the globally covered Global Navigation Satellite System (GNSS) signals. While GNSS signals have been widely used in remote sensing to monitor slow changes in the Earth’s surface, like sea level and snow depth, their ability to accurately detect highly dynamic target motions, such as human respiration, gestures, and intrusions, remains unclear. The main challenge arises from the interference brought by the large-scale satellite movement and severe GNSS signal errors. In this study, we present a novel GNSS signal enhancement system named FineSat to address these interference. Specifically, we first utilize polynomial representations to cancel satellite movement interference. Then, based on the analysis of GNSS signal errors, we propose a signal differential processing module to mitigate the errors. We implement our system on commercial devices and validate its performance in three sensing applications: respiration monitoring, gesture recognition, and intrusion detection. Results show that we achieve 0.42 bpm mean absolute error in respiration monitoring, 96.5% average accuracy in gesture recognition, and 98.6% accuracy in intrusion detection.

Large-scale trajectory data is crucial for applications like human mobility prediction and pandemic intervention. However, concerns over data privacy have limited access to real-world datasets. Advanced generative models offer a promising alternative for creating synthetic yet realistic trajectory data, but most existing methods focus on synchronous trajectories with fixed time intervals, which fail to capture the complexities of asynchronous trajectories, such as Point of Interest (POI) check-ins with uncertain intervals and varying lengths. To address this gap, we propose AutoSTDiff, a novel Autoregressive Spatio-Temporal denoising Diffusion model for asynchronous trajectory generation. AutoSTDiff includes two key components: (i) a hybrid embedding module that captures comprehensive spatio-temporal patterns considering human behavior and varying trajectory lengths, and (ii) a spatiotemporal diffusion model with a spatial status conversion module and a conditional spatio-temporal generation module for autoregressive trajectory generation. Extensive experiments on two public trajectory datasets show that AutoSTDiff outperforms state-of-the-art models, e.g., an increase of 51.2% and 43.8% on the length and G-rank metrics.

The Global Navigation Satellite System (GNSS) provides continuous high-precision positioning, enabling many applications such as vehicle navigation and pedestrian monitoring. However, in challenging environments such as urban canyons, positioning accuracy is significantly degraded due to multi-path and non-line-of-sight (NLOS) issues. To tackle this issue, we introduce a single point positioning (SPP) framework based on pseudorange correction, including three modules: feature extraction, pseudorange correction, and positioning model. Heavy pseudorange error is the primary cause of inaccurate localization, and in particular, we introduce GNSSFormer, a transformer-based model to obtain the pseudorange correction values. After preprocessing raw observations of GNSS, we designed a pseudorange correction model GNSSFormer containing a temporal transformer block and a multi-satellite joint spatial transformer block. GNSSFormer extracts multiple features related to both satellites and receivers, learning the complex global relationships between these features and the pseudorange errors to derive correction values. Subsequently, an extended Kalman filter (EKF)-based rauch tung striebel (RTS) smoothing SPP algorithm is employed to determine the location. Validation demonstrates that GNSSFormer significantly enhances SPP performance compared to state-of-the-art algorithms, improving positioning accuracy at least by 29.42% and 28.40% on two open source datasets and 18.91% on real-world dataset.

The delivery of goods to numerous indoor stores poses significant safety risks, with heavy, high-stacked packages on delivery trolleys posing a potential hazard to passersby. This paper reports our experiences of developing and operating DeMo, a practical system for real-time monitoring of indoor delivery. DeMo employs sensors attached to trolleys, utilizing Inertial Measurement Unit (IMU) and Bluetooth Low Energy (BLE) readings to detect delivery violations, such as speeding and the use of non-designated delivery paths, and ensure accurate matching of each delivery to its intended destination store. Unlike typical indoor localization applications, DeMo addresses unique challenges, including sensor placement and the complex electromagnetic characteristics encountered in underground settings. Specifically, DeMo adapts the classical logarithmic radio signal model to facilitate fingerprint-free localization, significantly reducing deployment and maintenance costs. DeMo has been operating since May 2020, covering more than 200 shops with 74,537 deliveries (6193.2 km) across 12 subway stations in Hong Kong. DeMo's 4-year operation witnessed a significant violation rate drop, from 19% (May 2020) to 0.9% (Mar 2024).

Individual cellular networks have been very robust to random cell tower failures due to redundant cell tower deployments. However, a large-scale clustered failure with multiple cell towers can lead to the loss of services of a cellular network. Recently, off-the-shelf smartphones can support multiple network standards, so cellular network infrastructure sharing is a promising direction to improve the service robustness under potential large-scale clustered tower failures. The existing work on cellular network robustness is usually limited to large-scale studies of individual networks or small-scale studies of multiple networks. In this work, we conduct the first investigation, to our knowledge, on cross-network infrastructure sharing benefits for enhancing robustness with afull cellular penetration rate. Our work is based on all cellular networks in Shenzhen, China, covering over 10 million cellular users. Specifically, we design a new metric to quantify cellular network robustness with or without cross-network sharing under both random and clustered cell tower failures. We further study the impact of different factors on robustness, including the number of networks, spatiotemporal dynamics, contextual factors, and a case study at two key transportation hubs. We provide a set of lessons learned based on our study, along with discussions of the results.

An address, a textual description of a geographical location, plays an important role in location-based services such as instant delivery. However, abnormal addresses (i.e., an address without detailed or accurate information) have led to significant costs. In real-world settings, abnormal address detection is not trivial because it needs to be completed in real-time to support massive online queries from mobile devices. In this study, we designFastAddr, a fast abnormal address detection framework, which detects abnormal addresses in real time.FastAddrconsists of a novel contrastive address augmentation module and a lightweight multi-head attention model. We further designFastAddr+ to enhanceFastAddrby utilizing large-scale spatial entities. A comprehensive three-phase evaluation is conducted. (i) We evaluateFastAddron a real-world dataset and it yields the average F1 of 85.7% in 0.058 milliseconds, which outperforms the state-of-the-art models by 47.4% with a similar detection time. (ii) An offline A/B test shows thatFastAddroutperforms the previous model significantly. (iii) We also conduct an online A/B test to compareFastAddrwith the deployed model, which shows an improvement of F1 by more than 20%. Moreover, we conduct two case studies on real industry data, demonstrating both the efficiency and effectiveness ofFastAddr.

Large language models (LLMs) have exhibited remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. Despite the recent success, current LLMs are not capable of processing complex audio information or conducting spoken conversations (like Siri or Alexa). In this work, we propose a multi-modal AI system named AudioGPT, which complements LLMs (i.e., ChatGPT) with 1) foundation models to process complex audio information and solve numerous understanding and generation tasks; and 2) the input/output interface (ASR, TTS) to support spoken dialogue. With an increasing demand to evaluate multi-modal LLMs of human intention understanding and cooperation with foundation models, we outline the principles and processes and test AudioGPT in terms of consistency, capability, and robustness. Experimental results demonstrate the capabilities of AudioGPT in solving 16 AI tasks with speech, music, sound, and talking head understanding and generation in multi-round dialogues, which empower humans to create rich and diverse audio content with unprecedented ease. Code can be found in https://github.com/AIGC-Audio/AudioGPT

Zhiqing Hong, Rongjie Huang, Xize Cheng, Yongqi Wang, Ruiqi Li, Fuming You, Zhou Zhao, Zhimeng Zhang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Note-level Automatic Singing Voice Transcription (AST) converts singing recordings into note sequences, facilitating the automatic annotation of singing datasets for Singing Voice Synthesis (SVS) applications.Current AST methods, however, struggle with accuracy and robustness when used for practical annotation.This paper presents ROSVOT, the first robust AST model that serves SVS, incorporating a multi-scale framework that effectively captures coarse-grained note information and ensures fine-grained frame-level segmentation, coupled with an attention-based pitch decoder for reliable pitch prediction.We also established a comprehensive annotation-and-training pipeline for SVS to test the model in realworld settings.Experimental findings reveal that ROSVOT achieves state-of-the-art transcription accuracy with either clean or noisy inputs.Moreover, when trained on enlarged, automatically annotated datasets, the SVS model outperforms its baseline, affirming the capability for practical application.Audio samples are available at https://rosvot.github.io.Codes can be found at https://github.com/RickyL- 2000/ROSVOT.

Multi-modal transportation leverages the advantages of various transportation modes, leading to more efficient urban traveling services. Accurately predicting transfer times between different modes provides guidance for tasks such as trip planning and transportation management. Most existing transfer time prediction works rely on strong assumptions, e.g., predetermined routes, assumed speeds, and predefined downstream transportation timetables. However, these assumptions are hard to hold in practice due to internal factors like individual preferences and external factors like dynamic traffic conditions. These factors are dynamic and vary with location and time, presenting a significant challenge. To address this, we introduce an adaptive transfer time prediction framework, AdaTrans, to forecast personalized transfer times between upstream and downstream transportation modes. Firstly, an attribute learning module is designed to model the trends of internal factors. Then a spatial-temporal adaptive learning component is designed to learn dynamic external factors. Finally, an aggregation component with a capsule network is employed to fuse the influences of these factors. The extensive evaluation results in two real-world datasets demonstrate that AdaTrans effectively harnesses insights from internal and external factors, outperforming state-of-the-art methods by ~20%.

Accurate prediction of order transportation time is essential for customer satisfaction in logistics. Existing methods based on origin-destination (OD) pairs do not consider the diversity of road segments, while route-based methods may fail to account for real-time traffic conditions due to the infrequent dispatch schedules of logistics vehicles. In reality, e-commerce platforms have collaborated with multiple logistics companies for parcel delivery, providing a richer dataset that offers a more comprehensive view of real-time transportation conditions. The key insight is that data from one company can serve as internal capability detectors and data from others can act as external environment detectors. However, a significant challenge arises in inferring travel-time-correlated station pairs across different companies, especially without full disclosure of station information. To address this, we design an Adaptive cross-platform Transportation time prediction framework built upon a hypergraph structure, named AdaTrans, comprising: i) A spatial-temporal routing graph learner employs node-centric and edge-centric hyperedges to address the complex, non-pairwise correlations among stations and station pairs within and across companies; ii) A spatial-temporal graph-based transportation time predictor that utilizes multi-task learning to enhance overall transportation time prediction by leveraging the correlations between interconnected sub-tasks (i.e., dwell and travel times prediction) Extensive evaluation with real-world data collected from JD.com, a leading e-commerce platform in China, demonstrates that consolidating records from other companies reduces RMSE, MAE, and MAPE by 12.63%, 5.18%, and 16.67%, compared to state-of-the-art methods.

Xiaoda Yang, Xize Cheng, Jiaqi Duan, Hongshun Qiu, Minjie Hong, Minghui Fang, Shengpeng Ji, Jialong Zuo, Zhiqing Hong, Zhimeng Zhang, Tao Jin. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Individual cellular networks have been very robust to random cell tower failure due to redundant cell tower deployments. However, a large-scale clustered failure (e.g., due to fiber cut or cyber attacks) with multiple cell towers can lead to the loss of services of a cellular network. Recently, off-the-shelf smartphones can support multiple network standards, so cellular network infrastructure sharing is a promising direction to improve the service robustness under potential large-scale clustered cell tower failure. The existing work on cellular network robustness is usually limited to large-scale studies of individual networks or small-scale studies of multiple networks. In this work, we conduct the first investigation, to our knowledge, into the benefits of cross-network infrastructure sharing for enhancing robustness at a full cellular penetration rate. We design a new metric to quantify cellular network robustness with or without cross-network sharing under both random and clustered cell tower failures. We further study the impact of spatial dynamics on cellular network robustness.

As location-based services (LBS) have grown in popularity, more human mobility data has been collected. The collected data can be used to build machine learning (ML) models for LBS to enhance their performance and improve overall experience for users. However, the convenience comes with the risk of privacy leakage since this type of data might contain sensitive information related to user identities, such as home/work locations. Prior work focuses on protecting mobility data privacy during transmission or prior to release, lacking the privacy risk evaluation of mobility data-based ML models. To better understand and quantify the privacy leakage in mobility data-based ML models, we design a privacy attack suite containing data extraction and membership inference attacks tailored for point-of-interest (POI) recommendation models, one of the most widely used mobility data-based ML models. These attacks in our attack suite assume different adversary knowledge and aim to extract different types of sensitive information from mobility data, providing a holistic privacy risk assessment for POI recommendation models. Our experimental evaluation using two real-world mobility datasets demonstrates that current POI recommendation models are vulnerable to our attacks. We also present unique findings to understand what types of mobility data are more susceptible to privacy attacks. Finally, we evaluate defenses against these attacks and highlight future directions and challenges.

Link prediction, determining if a relation exists between two entities, is an essential task in the analysis of heterogeneous graphs with diverse entities and relations. Despite extensive research in link prediction, most existing works focus on predicting the relation type between given pairs of entities. However, it is almost impractical to check every entity pair when trying to find most hidden relations in a billion-scale heterogeneous graph due to the billion squared number of possible pairs. Meanwhile, most methods aggregate information at the node level, potentially leading to the loss of direct connection information between the two nodes. In this paper, we introduce Paths2Pair, a novel framework to address these limitations for link prediction in billion-scale commercial heterogeneous graphs. (i) First, it selects a subset of reliable entity pairs for prediction based on relevant meta-paths. (ii) Then, it utilizes various types of content information from the meta-paths between each selected entity pair to predict whether a target relation exists. We first evaluate our Paths2Pair based on a large-scale dataset, and results show Paths2Pair outperforms state-of-the-art baselines significantly. We then deploy our Paths2Pair on JD Logistics, one of the largest logistics companies in the world, for business expansion. The uncovered relations by Paths2Pair have helped JD Logistics identify 108,709 contacts to attract new company customers, resulting in an 84% increase in the success rate compared to the state-of-the-practice solution, demonstrating the practical value of our framework. We have released the code of our framework at https://github.com/JQHang/Paths2Pair.

Recently, integrated warehouse and distribution logistics systems are widely used in E-commerce industries to adjust to constantly changing customer demands. It makes the prediction of purchase demand and delivery supply capacity a crucial problem to streamline operations and improve efficiency. The interaction between such demand and supply not only relies on their economic relationships but also on consumer psychology caused by daily events, such as epidemics, promotions, and festivals. Although existing studies have made great efforts in the joint prediction of demand and supply considering modeling the demand-supply interactions, they seldom refer to the impacts of diverse events. In this work, we propose MulSTE, a Multi-view Spatio-Temporal learning framework with heterogeneous Event fusion. Firstly, an Event Fusion Representation (EFR) module is designed to fuse the textual, numerical, and categorical heterogeneous information for emergent and periodic events. Secondly, a Multi-graph Adaptive Convolution Recurrent Network (MGACRN) is developed as the spatio-temporal encoder (ST-Encoder) to capture the evolutional features of demand, supply, and events. Thirdly, the Event Gated Demand-Supply Interaction Attention (EGIA) module is designed to model the demand-supply interactions during events. The evaluations are conducted on two real-world datasets collected from JD Logistics and public websites. The experimental results show that our method outperforms state-of-the-art baselines in various metrics.

Voice large language models (LLMs) cast voice synthesis as a language modeling task in a discrete space, and have demonstrated significant progress to date. Despite the recent success, the current development of voice LLMs in low-resource applications is hampered by data scarcity and high computational cost. In this work, we propose VoiceTuner, with a self-supervised pre-training and efficient fine-tuning approach for low-resource voice generation. Specifically, 1) to mitigate data scarcity, we leverage large-scale unlabeled dataset and pre-train VoiceTuner-SSL without pre-defined applications, which can be fine-tuned in downstream tasks; 2) to further reduce the high training cost in complete fine-tuning, we introduce a multiscale transformer adapter to effectively update only around 1% parameters as a plug-and-play module. Experimental results demonstrate that VoiceTuner-SSL presents strong acoustic continuations, and VoiceTuner achieves state-of-the-art results in rich-resource TTS evaluation compared with competitive baseline models. Low-resource (1h, 10h, 30h) downstream applications including zero-shot TTS, instruction TTS, and singing voice synthesis present VoiceTuner's superior audio quality and style similarity with reduced data requirement and computational cost. Audio samples are available at https://VoiceTuner.github.io

Yongqi Wang, Ruofan Hu, Rongjie Huang, Zhiqing Hong, Ruiqi Li, Wenrui Liu, Fuming You, Tao Jin, Zhou Zhao. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

The scarcity of high-quality and multi-task singing datasets significantly hinders the development of diverse controllable and personalized singing tasks, as existing singing datasets suffer from low quality, limited diversity of languages and singers, absence of multi-technique information and realistic music scores, and poor task suitability.To tackle these problems, we present GTSinger, a large Global, multi-Technique, free-to-use, high-quality singing corpus with realistic music scores, designed for all singing tasks, along with its benchmarks.Particularly,(1) we collect 80.59 hours of high-quality singing voices, forming the largest recorded singing dataset;(2) 20 professional singers across nine widely spoken languages offer diverse timbres and styles;(3) we provide controlled comparison and phoneme-level annotations of six commonly used singing techniques, helping technique modeling and control;(4) GTSinger offers realistic music scores, assisting real-world musical composition;(5) singing voices are accompanied by manual phoneme-to-audio alignments, global style labels, and 16.16 hours of paired speech for various singing tasks.Moreover, to facilitate the use of GTSinger, we conduct four benchmark experiments: technique-controllable singing voice synthesis, technique recognition, style transfer, and speech-to-singing conversion.

Node ranking in heterogeneous graphs, which quantifies the relative importance of nodes, can often be improved by incorporating information from relevant paths. Graph database and heterogeneous graph neural network (HGNN) are two main approaches to better solve this problem. Graph databases support efficient path queries for flexible path types but require manual design to combine results for node ranking. Conversely, current HGNNs can automatically integrate semantic information from multiple linear path types for accurate node ranking. However, our experiments show that they fail to outperform a multi-layer perceptron model that utilizes features extracted from multiple nonlinear conditional paths, which can be handled by graph databases. Therefore, we aim to enable HGNN to take advantage of these path types for better performance. However, HGNNs require a generalized path schema to define the structure of input paths, and incorporating each additional path type will significantly increase the required system memory and sampling time for HGNNs. To address these limitations, we introduce CompNode, a novel framework based on a new unified path schema definition called Complex-path, which is used to describe all the required path types, including nonlinear conditional path types. Then, we design a pre-aggregation method to reduce the required system memory and sampling time by pre-aggregating the same type of complex-path. Furthermore, we develop a model that combines semantic information from all aggregated complex-paths for accurate node ranking. Real-world experiments on identifying top potential high-value customers show CompNode outperforms state-of-the-art HGNNs by 20% in average precision and the previously deployed graph database method by 252% in success rate.

Geocoding, associating textual addresses with corresponding GPS coordinates, is vital for many location-based services (e.g., logistics, ridesharing, and social networks). One of the most common Geocoding solutions is using commercial map services such as Google Maps. However, this is typically not practical for some location-based service providers due to real-world challenges like commercial competition and high costs (recurring fees). In this paper, we design a new cost-effective Geocoding framework to automatically infer the geographic coordinates from textual addresses. To achieve this, we take the E-Commerce logistics service as a concrete scenario and designCoMiner, an unsupervised coordinate inference framework based on textual address data, delivery event data, and courier trajectory data.CoMinerincludes three main components, (1) A POI-level clustering model, (2) A Delivery Mobility Graph (DMG), and (3) A behavior-driven address ranking model. Furthermore, we designCoMiner-W, a coordinates mining algorithm based on WiFi data, to further enhance the effectiveness ofCoMiner. We conduct extensive experiments on three large-scale datasets whereCoMineroutperforms the state-of-the-art methods by 20.3%. Moreover, we have designed an abnormal delivery event detection system based onCoMinerand deployed it at JD Logistics, which brings a significant reduction in abnormal delivery event rates.

Fine-grained community-building information, such as building names and accurate geographical coordinates, is critical for a range of practical applications like navigation and door-to-door services (e.g., on-demand delivery and last-mile delivery). A common practice of traditional methods to gather community-building information usually relies on manual collection, which is typically labor-intensive and time-consuming. To address these issues, we utilize the massive data generated from e-commerce delivery services and design a framework, AutoBuild, for fine-grained large-scale community-building labeling. AutoBuild consists of two main components: (i) a Location Candidate Detection Module that identifies potential building names and coordinates from multi-source delivery data, and (ii) a Progressive Building Matching Model that employs trajectory modeling, human behavior analysis, and heterogeneous graph alignment to match building names and coordinates. To evaluate the performance of AutoBuild, we applied it to two real-world multi-modal datasets from Beijing City and Chengdu City. The results reveal that AutoBuild significantly outperforms multiple baseline models by 50-meter accuracy of 81.8% and 100-meter accuracy of 95.9% in Beijing City. More importantly, we conduct a real-world case study to demonstrate the practical impact of AutoBuild in last-mile delivery.

Points of Interest (POIs), such as entertainment, dining, and living, are crucial for urban planning and location-based services. However, the high dynamics and expensive updating costs of POIs pose a key roadblock for their urban applications. This is especially true for developing countries, where active economic activities lead to frequent POI updates (e.g., merchants closing down and new ones opening). Therefore, POI updating, i.e., detecting new POIs and different names of the same POIs (alias) to update the POI database, has become an urgent but challenging problem to address. In this paper, we attempt to answer the research question of how to detect and update large-scale POIs via a low-cost approach. To do so, we propose a novel framework called UrbanPOI, which formulates the POI updating problem as a tagging and detection problem based on multi-modal logistics delivery data. UrbanPOI consists of two key modules: (i) a hierarchical POI candidate generation module based on the POINet model that detects POIs from shipping addresses; and (ii) a new POI detection module based on the Siamese Attention Network that models multi-modal data and crowd intelligence. We evaluate our framework on real-world logistics delivery datasets from two Chinese cities. Extensive results show that our model outperforms state-of-the-art models in Beijing City by 26.2% in precision and 10.7% in F1-score, respectively.

Last-mile delivery from delivery stations to customers’ places is now mainly finished by dedicated couriers. In practice, each courier generally collects orders destined for one delivery area at the delivery station and delivers orders to customers. However, the long distance between the delivery station and the delivery area due to practical reasons, e.g., expensive delivery station rental fee in the downtown area, increases the delivery courier’s travel time and decreases the efficiency of the state-of-the-practice last-mile delivery scheme. In this paper, we solve the problem with relay transportation, where a relay courier collects orders at the delivery station and sends them to delivery couriers, and delivery couriers focus on the order delivery at corresponding delivery areas. We design a real-time relay courier scheduling system called REDE to minimize the average relay order delivery time (ARODT) considering the relay and delivery couriers’ mobility and the order destination distribution. First, a heterogeneous task aware route prediction algorithm is proposed to characterize the delivery courier’s mobility. Then a distance-aware greedy algorithm and an ARODT-constrained exchange algorithm are designed to generate the relay route, which is updated with real-time order pickup requests. Extensive evaluation results with real-world logistics data from 100 delivery stations in 38 cities show that REDE reduces ARODT by up to 8.4% compared to baseline methods. The online A/B tests show that compared to the state-of-the-practice method, REDE improves the delivery courier’s working efficiency and the daily number of pickup orders by 20.13% and 4.51%, respectively.

Though previous works have shown remarkable achievements in singing voice generation, most existing models focus on one specific application and there is a lack of unified singing voice synthesis models. In addition to low relevance among tasks, different input modalities are one of the most intractable hindrances. Current methods suffer from information confusion and they can not perform precise control. In this work, we propose UniSinger, a unified end-to-end singing voice synthesizer, which integrates three abilities related to singing voice generation: singing voice synthesis (SVS), singing voice conversion (SVC), and singing voice editing (SVE) into a single framework. Specifically, we perform representation disentanglement for controlling different attributes of the singing voice. We further propose a cross-modality information matching method to close the distribution gap between multi-modal inputs and achieve end-to-end training. The experiments conducted on the OpenSinger dataset demonstrate that UniSinger achieves state-of-the-art results in three applications. Further extensive experiments verify the capability of representation disentanglement and information matching, reflecting that UniSinger enjoys great superiority in sample quality, timbre similarity, and multi-task compatibility. Audio samples can be found in https://unisinger.github.io/Samples/.

The popularity of e-commerce has promoted the rapid development of the logistics industry in recent years. As an important step in logistics, last-mile delivery from delivery stations to customers' addresses is now mainly finished by couriers, which requires accurate workload assessment based on actual efforts. However, the state-of-the-practice assessment methods neglect a vital factor that orders with the same customer's address (i.e., Homogeneous orders) can be delivered in a group (i.e., in a single trip) or separately (i.e., in multiple trips). It would cause unfair assessment among couriers if following the same rule. Thus, grouping homogeneous order accurately in the workload assessment is significant for achieving fair courier's workload assessment. To this end, we design, implement, and deploy a nationwide homogeneous order grouping system called FHOG for improving the accuracy of homogeneous order grouping in last-mile delivery for fair courier's workload assessment. FHOG utilizes the courier's reporting behavior for order inspection, collection, and delivery to identify homogeneous orders in the delivery station simultaneously for homogeneous order grouping. Compared with the state-of-the-practice method, our evaluation shows FHOG can effectively reduce order amounts with the higher and lower assessed courier's workload. We further deploy FHOG online in 8336 delivery stations to provide homogeneous order grouping service for more than 120 thousand couriers and 12 million daily orders. The results of the two surveys show that the couriers' acceptance rate is improved by 67% with FHOG after the promotion.

Geocoding, associating textual addresses with corresponding GPS coordinates, is vital for many location-based services (e.g., logistics, ridesharing, and social networks). One of the most common Geocoding solutions is using commercial map services (e.g., Google Maps) by uploading textual addresses to obtain corresponding coordinates. However, this is typically not practical for some location-based service providers due to real-world challenges like commercial competition and high costs (recurring fees). In this paper, we design a new cost-effective Geocoding framework to automatically infer the geographic coordinates from textual addresses for service providers. To achieve this, we take the E-Commerce logistics service as a concrete scenario and design CoMiner, an unsupervised coordinate inference framework based on textual address data, delivery event data, and courier trajectory data. There are three main components in CoMiner. (1) A POI-level clustering model by modeling customers' shopping patterns at different spatial granularities; (2) A Delivery Mobility Graph (DMG) by modeling couriers' delivery events and geographic coordinates; (3) A behavior-driven address ranking model by mining couriers' uncertain reporting behaviors to further infer coordinates on DMG. We extensively verify the performance of CoMiner with a three-phase evaluation from data-driven experiments to real-world deployment. (i) We conduct extensive experiments on three large-scale datasets where CoMiner achieves an average accuracy of 95.1%, which outperforms the state-of-the-art methods by 20.3%. (ii) We deploy CoMiner in JD Logistics, inferring coordinates for over 30 million addresses with an average accuracy of 93.3%. (iii) We utilize CoMiner for two Geocoding-based applications, i.e., parcel re-routing optimization and abnormal delivery event detection.

An address, a textual description of a physical location, plays an important role in location-based services such as on-demand delivery and e-commerce. However, abnormal addresses (i.e., an address without detailed information representing a spatial location) have led to significant costs. In real-world settings like e-commerce, abnormal address detection is not trivial because it needs to be completed in real-time to support massive online queries. In this study, we design FastAddr, a fast abnormal address detection framework, which detects abnormal addresses among millions of addresses in a short time. By investigating and modeling the hierarchical structure of address data, we first design a novel contrastive address augmentation approach to generate training data via learning the entity transition probability matrix. We further design a lightweight multi-head attention model for learning compact address representation by modeling the address characteristics. We conduct a comprehensive three-phase evaluation. (i) We evaluate FastAddr on a real-world dataset and it yields the average F1 of 85.7% in 0.058 milliseconds, which outperforms the state-of-the-art models by 47.4% with similar detection time. (ii) An offline A/B test shows that FastAddr outperforms the previous deployed model significantly. (iii) We also conduct an online A/B test to compare FastAddr with the deployed model, which shows an improvement of F1 by more than 20%. Moreover, a real-world case study demonstrates both the efficiency and effectiveness of FastAddr.

Electroencephalogram (EEG) signals for motor imagery (MI) are easily influenced by the environment and the state of the subject, which exhibit temporal and spatial variance. And this variance is more significant across subjects and sessions, which imposes limitations on the cross-domain MI tasks. To address this problem, we propose a Multi-level Spatial-Temporal Adaptation Network (MSTAN), extracting domain-invariant multi-level spatial-temporal features to overcome domain differences. First, stacked spatial-temporal graph convolution (STGCN) layers and an attention-based readout module are designed to extract spatial-temporal patterns of EEGs at multiple levels. An adaptation scheme is then introduced to narrow domain differences: 1) Individual graph parameters for the source and target domains are designed at each STGCN layer to capture the domain-specific brain region dynamic relationships; 2) The differences of spatial-temporal features between the source and target domain are reduced by minimizing the distribution distance. Experiments are conducted to evaluate the proposed method on a public dataset and the results show that our method achieves state-of-the-art performance in cross-domain motor imagery classification.

The availability of large amounts of data has driven the fast development of many research fields such as computer vision. However, the sharing of location data has been limited due to the concern of data privacy. Cellular location data contains user location information which reflects human mobility patterns. Therefore, in this study, we propose a novel Generative Adversarial Network (GAN) and apply the model to generate cellular location data as a case study for location-based sensing applications. The key insight of this study is that individual mobility correlates with user-specific information, e.g., age, gender. Therefore, to better capture the underlying pattern of human mobility, we design a soft-label conditional GAN which utilizes user-specific information to generate individual movement trajectories. This work plans to train a generator on a large real-world cellular location dataset and evaluate the synthetic data in terms of both utility and privacy.