Tax code prediction is a crucial yet underexplored task in automating invoicing and compliance management for large-scale e-commerce platforms. Each product must be accurately mapped to a node within a multi-level taxonomic hierarchy defined by national standards, where errors lead to financial inconsistencies and regulatory risks. This paper presents Taxon, a semantically aligned and expert-guided framework for hierarchical tax code prediction. Taxon integrates (i) a feature-gating mixture-of-experts architecture that adaptively routes multi-modal features across taxonomy levels, and (ii) a semantic consistency model distilled from large language models acting as domain experts to verify alignment between product titles and official tax definitions. To address noisy supervision in real business records, we design a multi-source training pipeline that combines curated tax databases, invoice validation logs, and merchant registration data to provide both structural and semantic supervision. Extensive experiments on the proprietary TaxCode dataset and public benchmarks demonstrate that Taxon achieves state-of-the-art performance, outperforming strong baselines. Further, an additional full hierarchical paths reconstruction procedure significantly improves structural consistency, yielding the highest overall F1 scores. Taxon has been deployed in production within Alibaba's tax service system, handling an average of over 500,000 tax code queries per day and reaching peak volumes above five million requests during business event with improved accuracy, interpretability, and robustness.

User-item interaction histories are pivotal for sequential recommendation systems but often include noise, such as unintended clicks or actions that fail to reflect genuine user preferences. To address this, we propose Learned Item Shortcuts for Sequential Recommendation (LISRec), a novel framework that explicitly captures stable preferences by extracting personalized semantic shortcuts from historical interactions. LISRec first learns task-agnostic semantic representations to assess item similarities, then constructs a personalized semantic graph over all user-interacted items. By identifying the maximal semantic connectivity subset within this graph, LISRec selects the most representative items as semantic shortcuts to guide user preference modeling. This focused representation filters out irrelevant actions while preserving the diversity of genuine interests. Experimental results on the Yelp and Amazon Product datasets illustrate that LISRec achieves a 13% improvement over baseline recommendation models, showing its effectiveness in capturing stable user interests. Further analysis indicates that shortcut-based histories better capture user preferences, making more accurate and relevant recommendations. All codes and datasets are available at https://github.com/NEUIR/LISRec.

Conversion objectives in large-scale recommender systems are sparse, making them difficult to optimize. Generative recommendation (GR) partially alleviates data sparsity by organizing multi-type behaviors into a unified token sequence with shared representations, but conversion signals remain insufficiently modeled. While recent behavior-aware GR models encode behavior types and employ behavior-aware attention to highlight decision-related intermediate behaviors, they still rely on standard attention over the full history and provide no additional supervision for conversions, leaving conversion sparsity largely unresolved. To address these challenges, we propose RCLRec, a reverse curriculum learning–based GR framework for sparse conversion supervision. For each conversion target, RCLRec constructs a short curriculum by selecting a subsequence of conversion-related items from the history in reverse. Their semantic tokens are fed to the decoder as a prefix, together with the target conversion tokens, under a joint generation objective. This design provides additional instance-specific intermediate supervision, alleviating conversion sparsity and focusing the model on the user's critical decision process. We further introduce a curriculum quality-aware loss to ensure that the selected curricula are informative for conversion prediction. Experiments on offline datasets and an online A/B test show that RCLRec achieves superior performance, with +2.09% advertising revenue and +1.86% orders in online deployment.

To maintain the company's talent pool, recruiters need to continuously search for resumes from third-party websites (e.g., LinkedIn, Indeed). However, fetched resumes are often incomplete and inaccurate. To improve the quality of third-party resumes and enrich the company's talent pool, it is essential to conduct duplication detection between the fetched resumes and those already in the company's talent pool. Such duplication detection is challenging due to the semantic complexity, structural heterogeneity, and information incompleteness of resume texts. To this end, we propose MHSNet, an multi-level identity verification framework that fine-tunes BGE-M3 using contrastive learning. With the fine-tuned BGE-M3, MHSNet generates multi-level sparse and dense representations for resumes, enabling the computation of corresponding multi-level semantic similarities. Moreover, the state-aware Mixture-of-Experts (MoE) is employed in MHSNet to handle diverse incomplete resumes. Experimental results verify the effectiveness of MHSNet.

Transformer-based sequential recommendation (TSR) models have shown superior performance in recommendation systems, where the quality of item representations plays a crucial role. Classical representation methods integrate item features using concatenation or neural networks to generate homogeneous representation sequences. While straightforward, these methods overlook the heterogeneity of item features, limiting the transformer's ability to capture fine-grained patterns and restricting scalability. Recent studies have attempted to integrate user-side heterogeneous features into item representation sequences, but item-side heterogeneous features, which are vital for performance, remain excluded. To address these challenges, we propose a Heterogeneous Information Transformer model for Sequential Recommendation (HeterRec), which incorporates Heterogeneous Token Flatten Layer (HTFL) and Hierarchical Causal Transformer Layer (HCT). Our HTFL is a novel item tokenization method that converts items into a heterogeneous token set and organizes these tokens into heterogeneous sequences, effectively enhancing performance gains when scaling up the model. Moreover, HCT introduces token-level and item-level causal transformers to extract fine-grained patterns from the heterogeneous sequences. Experiments on offline and online datasets show that the HeterRec model achieves superior performance.

With the rapid growth of user historical behavior data, user interest modeling has become a prominent aspect in Click-Through Rate (CTR) prediction, focusing on learning user intent representations. However, this complexity poses computational challenges, requiring a balance between model performance and acceptable response times for online services. Traditional methods often utilize filtering techniques. These techniques can lead to the loss of significant information by prioritizing top K items based on item attributes or employing low-precision attention mechanisms. In this study, we introduce LREA, a novel attention mechanism that overcomes the limitations of existing approaches while ensuring computational efficiency. LREA leverages low-rank matrix decomposition to optimize runtime performance and incorporates a specially designed loss function to maintain attention capabilities while preserving information integrity. During the inference phase, matrix absorption and pre-storage strategies are employed to effectively meet runtime constraints. The results of extensive offline and online experiments demonstrate that our method outperforms state-of-the-art approaches.

Intellectual Property (IP) management involves strategically protecting and utilizing intellectual assets to enhance organizational innovation, competitiveness, and value creation. Patent matching is a crucial task in intellectual property management, which facilitates the organization and utilization of patents. Existing models often rely on the emergent capabilities of Large Language Models (LLMs) and leverage them to identify related patents directly. However, these methods usually depend on matching keywords and overlook the hierarchical classification and categorical relationships of patents. In this paper, we propose MemGraph, a method that augments the patent matching capabilities of LLMs by incorporating a memory graph derived from their parametric memory. Specifically, MemGraph prompts LLMs to traverse their memory to identify relevant entities within patents, followed by attributing these entities to corresponding ontologies. After traversing the memory graph, we utilize extracted entities and ontologies to improve the capability of LLM in comprehending the semantics of patents. Experimental results on the PatentMatch dataset demonstrate the effectiveness of MemGraph, achieving a 17.68% performance improvement over baseline LLMs. The further analysis highlights the generalization ability of MemGraph across various LLMs, both in-domain and out-of-domain, and its capacity to enhance the internal reasoning processes of LLMs during patent matching. All data and codes are available at https://github.com/NEUIR/MemGraph.

Temporal Knowledge Graph (TKG) extrapolation aims to predict future missing facts based on historical information, which has exhibited both semantics and topology of events. The mainstream methods have advanced the prediction performance by exploring the potential of topology representations of TKGs based on dedicated temporal Graph Neural Networks (GNNs). Until recently, few Language Models (LM) based methods have attempted to model the semantic representations of TKGs, however, lacking specific designs for the topology information. Therefore, we propose a Semantic TOpology REpresentation learning (STORE) framework enhanced by LMs to bridge the gap between the semantics and topology of TKGs. Firstly, we tackle the challenge of long historical facts modeling by a time-aware sampling based on semantic priors to extract concise yet precise facts. Secondly, we handle the challenge of the interaction between topology and semantics by transforming graph representations into virtual tokens that are then integrated with generated prompts and fed into LMs. Finally, multi-head attention is adopted to obtain better semantic topology representations, thereby achieving joint optimization of both temporal GNNs and LMs. Extensive experiments on five datasets show that our STORE outperforms state-of-the-art GNNs- and LM-based methods.

Dynamic pricing, which suggests the optimal prices based on the dynamic demands, has received considerable attention in academia and industry. On online hotel booking platforms, room demand fluctuates due to various factors, notably hotel popularity and competition. In this paper, we propose a dynamic pricing approach with popularity and competitiveness-aware demand learning. Specifically, we introduce a novel demand function that incorporates popularity and competitiveness coefficients to comprehensively model the price elasticity of demand. We develop a dynamic demand prediction network that focuses on learning these coefficients in the proposed demand function, enhancing the interpretability and accuracy of price suggestion. The model is trained in a multi-task framework that effectively leverages the correlations of demands among groups of similar hotels to alleviate data sparseness in room-level occupancy prediction. Comprehensive experiments conducted on real-world datasets validate the superiority of our method over state-of-the-art baselines in both demand prediction and dynamic pricing. Our model has been successfully deployed on a popular online travel platform, serving tens of millions of users and hoteliers.

Train transfer plan ranking has become the core business of online travel platforms (OTPs), due to the flourish development of high- speed rail technology and convenience of booking trains online. Currently, mainstream OTPs adopt rule-based or simple preference- based strategies to rank train transfer plans. However, the insuf- ficient emphasis on the costs of plans and the negligence of con- sidering reference transfer plans make these existing strategies less effective in solving the personalized ranking problem of train transfer plans. To this end, a novel personalized deep network (Plan- Ranker) is presented in this paper to better address the problem. In PlanRanker, a personalized learning component is first proposed to capture both of the query semantics and the target transfer plan- relevant personalized interests of a user over the user's behavior log data. Then, we present a cost learning component, where both of the price cost and the time cost of a target transfer plan are emphasized and learned. Finally, a reference transfer plan learning component is designed to enable the whole framework of PlanRanker to learn from reference transfer plans which are pieced together by plat- form users and thus reflect the wisdom of crowd. PlanRanker is now successfully deployed at Alibaba Fliggy, one of the largest OTPs in China, serving millions of users every day for train ticket reservation. Offline experiments on two production datasets and a country-scale online A/B test at Fliggy both demonstrate the superiority of the proposed PlanRanker over baselines.

In the classical e-commerce platforms, the personalized product-tying recommendation has proven to be of great added value, which improves users' purchase willingness to product-tying by displaying the suitable marketing creative. In this paper, we present a new recommendation problem, i.e., the Pop-up One-time Marketing (POM), where the product-tying marketing creative only pops up one time when the user pays for the main item. POM has become a ubiquitous application in e-commerce platforms, e.g., buy the mobile tying mobile case and buy flight ticket tying insurance. However, many existing recommendation methods are sub-optimal for the creative marketing in the POM scenario due to unconsidering the unique characteristics in the scenario. To tackle this problem, we propose a novel framework named Event-aware Adaptive Clustering Uplift Network (EACU-Net) for the POM scenario, which is to our best knowledge the first attempt along this line. EACU-Net contains three modules: (1) the event-aware graph cascading learning, which employs a heterogeneous graph network to comprehensively learn the embedding for the user attributes, event categories, and creative elements by stage. (2) an adaptive clustering uplift network, which learns the sensitivity of users to creatives under the same context. (3) an event-aware information gain network to learn more information from samples with event affection. Extensive offline and online evaluations on a real-world e-commerce platform demonstrate the superior performance of the proposed model compared with the state-of-the-art method.

Personalized recommendation has been instrumental in many real applications. Despite the great progress, the underlying multi-scenario characteristics (e.g., users may behave differently under different scenarios) are largely ignored by existing recommender systems. Intuitively, modeling different scenarios properly could significantly improve the recommendation accuracy, and some existing work has explored this direction. However, these work assumes the scenarios are explicitly given, and thus becomes less effective when such information is unavailable. To complicate things further, proper scenario modeling from data is challenging and the recommendation models may easily overfit to some scenarios. In this paper, we propose a multi-scenario learning framework, MUSENET, for personalized recommendation. The key idea of MUSENET is to learn multiple implicit scenarios from the user behaviors, with a careful design inspired by the causal interpretation of recommender systems to avoid the overfitting issue. Additionally, since users' repeat consumptions account for a large part of the user behavior data on many e-commerce platforms, a repeat-aware mechanism is integrated to handle users' repurchase intentions within each scenario. Comprehensive experimental results on both industrial and public datasets demonstrate the effectiveness of the proposed approach compared with the state-of-the-art methods.

Graph learning-based collaborative filtering (GLCF), which is built upon the message-passing mechanism of graph neural networks (GNNs), has received great recent attention and exhibited superior performance in recommender systems. However, although GNNs can be easily compromised by adversarial attacks as shown by the prior work, little attention has been paid to the vulnerability of GLCF. Questions like can GLCF models be just as easily fooled as GNNs remain largely unexplored. In this article, we propose to study the vulnerability of GLCF. Specifically, we first propose an adversarial attack against CLCF. Considering the unique challenges of attacking GLCF, we propose to adopt the greedy strategy in searching for the local optimal perturbations and design a reasonable attacking utility function to handle the non-differentiable ranking-oriented metrics. Next, we propose a defense to robustify GCLF. The defense is based on the observation that attacks usually introduce suspicious interactions into the graph to manipulate the message-passing process. We then propose to measure the suspicious score of each interaction and further reduce the message weight of suspicious interactions. We also give a theoretical guarantee of its robustness. Experimental results on three benchmark datasets show the effectiveness of both our attack and defense.

Fliggy of Alibaba group is one of the largest online travel platform (OTPs) in China, which provides travel products and travel experiences for tens of millions of online users by the personalized recommendation system (RS). User's future travel intent prediction is one key problem in travel scenario, which decides where and what to recommend, e.g., traveling to a surrounding city or a distant city. Such travel intent prediction problem has a lot of important applications, e.g., to push a notification with surrounding scenic spots recommendation to a user with intent to travel around, or to enable personalized promotion strategies to users with different intents. Existing studies on user's intent are largely sub-optimal for users' travel intent prediction at OTPs, since they rarely pay attentions to the characteristics of the travel industry, namely, user behavior sparsity due to low frequency of travel, spatial-temporal periodicity patterns, and the correlations between user's online and offline behaviors. In this paper, to address these challenges, we propose a dual channel intent evolution network based online-offline periodicity-aware network, DCIEN, for user's future travel intent prediction. In particular, it consists of two basic components including 1) Spatial-temporal Intent Patterns Network(ST-IPN), which exploits users' periodic intent patterns from offline data based on convolutional neural networks; 2) Periodicity-aware Intent Evolution Network(PA-IEN), which captures user's instant intent from online behaviors data and the interactions between online and offline intents. Extensive offline and online experiments on a real-world OTP demonstrate the superior performance of DCIEN over state-of-the-art methods.

Sequential matching, which aims to predict the item a user will next interact with in the sequential context of the user's historical behaviors, is widely adopted in recommender systems. Existing works mainly characterize the sequential context as the dependencies of user interactions, which is less effective for online travel recommendation where users' behaviors are highly correlated with theirstages in the travel life cycle. Specifically, users on an online travel platform (OTP) usually go through different stages (e.g., exploring a destination, planning an itinerary), and make several correlated interactions (e.g., booking a flight, reserving a hotel, renting a car) at each stage. In this paper, we propose to capture the deep sequential context by modeling the evolving of user stages, and develop a novel stage-aware deep sequential matching network (SASNet) that incorporates inter-stage and intra-stage dependencies over stage-augmented interaction sequence for more accurate and interpretable recommendation. Extensive experiments on real-world datasets validate the superiority of our model for both online travel recommendation and general next-item recommendation. Our model has been successfully deployed at Fliggy, one of the most popular OTPs in China, and shows good performance in serving online traffic.

In this paper, we propose a novel elastic demand function that captures the price elasticity of demand in hotel occupancy prediction. We develop a price elasticity prediction model (PEM) with a competitive representation module and a multi-sequence fusion model to learn the dynamic price elasticity from a complex set of affecting factors. Moreover, a multi-task framework consisting of room- and hotel-level occupancy prediction tasks is introduced to PEM to alleviate the data sparsity issue. Extensive experiments on real-world datasets show that PEM outperforms other state-of-the-art methods for both occupancy prediction and dynamic pricing. PEM model has been successfully deployed at Fliggy and shown good performance in online hotel booking services.

Price, a crucial factor determining whether a user will purchase an item, has attracted considerable attention in personalized ranking and recommendation. Existing studies commonly assume that only item price affects user online purchase decisions. However, in reality, users not only focus on the price of an item itself but also compare the price with the item's “comparison prices,” including its past prices, prices of similar items, and prices on other e-commerce platforms. Without carefully considering these comparison prices, methods fail to capture the purchase motivation attributable to prices comprehensively. To address this problem, in this paper, we introduce the concept of item price competitiveness. An item's price competitiveness measures the advantage of the item's price over its comparison prices. Then, a novel Price Competitiveness-aware Network (PCNet) is proposed to predict users' purchase behaviors by explicitly considering the price competitiveness of items. Specifically, PCNet consists of three key modules, and each module exploits one corresponding facet of price competitiveness. We leverage prior knowledge discovered from a real-world dataset to guide module designs, thus enhancing the performance and interpretability of the PCNet. Offline experiments show the superiority of the PCNet and verify the effectiveness of each module. Moreover, PCNet has been deployed online in a hotel search engine at Fliggy and benefits both the platform and users.

Origin-Destination recommendation that recom-mends personalized origin city (O) and destination city (D) of flight itinerary is of great value for both Online Travel Platforms (OTPs) and users. Existing studies on next location recommendation propose to model the sequential regularity of users' check-in location sequences, but cannot well solve two new challenges facing OTPs, namely the necessity of exploring O&D and learning O&D as a whole. To this end, we propose a novel personalized Origin-Destination ranking NETwork (ODNET) for flight recommendation. In particular, a heterogeneous spatial graph (HSG) which models historical interactions between users and cities is designed at first. HSG is then deployed in ODNET to identify user preference Os and Ds by exploring the neighbor-hood information in HSG. To cope with the second challenge, the idea of multi-task learning is employed by ODNET to learn$O$and$D$jointly so as to capture their correlations. Moreover, temporal information of Os and Ds are also considered to further improve the accuracy of origin-destination recommendation. An offline experiment on multiple real-world datasets and an online A/B test both show the superiority of ODNET towards the state-of-the-art methods. Further, the implementation and deployment details of the proposed ODNET at Fliggy, one of the most popular OTPs in China, are also described. ODNET has now been successfully applied to provide high-quality flight recommendation service at Fliggy, serving tens of millions of users.

Hotel search ranking is the core function of Online Travel Platforms (OTPs), while geography information of location entities involved in it plays a critically important role in guaranteeing its ranking quality. The closest line of works to the hotel search ranking problem is thus the next POI (or location) recommendation problem, which has extensive works but fails to cope with two new challenges, i.e., consideration of two more location entities and effective utilization of geographical information, in a hotel search ranking scenario. To this end, we propose a General Geography-aware representation NETwork (G2NET for short) to better represent geography information of location entities so as to optimize the hotel search ranking. In G2NET, to address the first challenge, we first propose the concept of Geography Interaction Schema (GIS) which is a meta template for representing the arbitrary number of location entity types and their interactions. Then, a novel geography interaction encoder is devised providing general representation ability for an instance of GIS, followed by an attentive operation that aggregates representations of instances corresponding to all historically interacted hotels of a user in a weighted manner. The second challenge is handled by the combined application of three proposed geography embedding modules in G2NET, each of which focuses on computing embeddings of location entities based on a certain aspect of geographical information of location entities. Moreover, a self-attention layer is deployed in G2NET, to capture correlations among historically interacted hotels of a user which provides non-trivial functionality of understanding the user's behaviors. Both offline and online experiments show that G2NET outperforms the state-of-the-art methods. G2NET has now been successfully deployed to provide the high-quality hotel search ranking service at Fliggy, one of the most popular OTPs in China, serving tens of millions of users.

Online travel platforms (OTPs), e.g., booking.com and Ctrip.com, deliver travel experiences to online users by providing travel-related products. One key problem facing OTPs is to predict users' future travel destination, which has many important applications, e.g., proactively recommending users flight tickets or hotels in the destination city. Although much progress has been made for the next POI recommendation, they are largely sub-optimal for travel destination prediction on OTPs, due to the unique characteristics exhibited from users' travel behaviors such as offline spatial-temporal periodicity and online multi-interest exploration. In this paper, we propose an online-offline periodicity-aware information gain network, OOPIN, for travel destination prediction on OTPs. The key components of the model are (1) an offline mobility pattern extractor, which extracts spatial-temporal periodicity along with the sequential dependencies from the visited city sequence; and (2) an online multi-interests exploration module that discovers destinations that the user might be interested in but not yet visited from their online interaction data.Comprehensive experiments on real-world OTP demonstrate the superior performance of the proposed model for travel destination prediction compared with state-of-the-art methods.

In many classical e-commerce platforms, personalized recommendation has been proven to be of great business value, which can improve user satisfaction and increase the revenue of platforms. In this paper, we present a new recommendation problem, Trigger-Induced Recommendation (TIR), where users’ instant interest can be explicitly induced with a trigger item and follow-up related target items are recommended accordingly. TIR has become ubiquitous and popular in e-commerce platforms. In this paper, we figure out that although existing recommendation models are effective in traditional recommendation scenarios by mining users’ interests based on their massive historical behaviors, they are struggling in discovering users’ instant interests in the TIR scenario due to the discrepancy between these scenarios, resulting in inferior performance. To tackle the problem, we propose a novel recommendation method named Deep Interest Highlight Network (DIHN) for Click-Through Rate (CTR) prediction in TIR scenarios. It has three main components including 1) User Intent Network (UIN), which responds to generate a precise probability score to predict user’s intent on the trigger item; 2) Fusion Embedding Module (FEM), which adaptively fuses trigger item and target item embeddings based on the prediction from UIN; and (3) Hybrid Interest Extracting Module (HIEM), which can effectively highlight users’ instant interest from their behaviors based on the result of FEM. Extensive offline and online evaluations on a real-world e-commerce platform demonstrate the superiority of DIHN over state-of-the-art methods. Our code is available 1.

The travel marketing platform of Alibaba serves an indispensable role for hundreds of different travel scenarios from Fliggy, Taobao, Alipay apps, etc. To provide personalized recommendation service for users visiting different scenarios, there are two critical issues to be carefully addressed. First, since the traffic characteristics of different scenarios, e.g., individual data scale or representative topic, are significantly different, it is very challenging to train a unified model to serve all. Second, during the promotion period, the exposure of some specific items will be re-weighted due to manual intervention, resulting in biased logs, which will degrade the ranking model trained using these biased data. In this paper, we propose a novel Scenario-Aware Ranking Network (SAR-Net) to address these issues. SAR-Net harvests the abundant data from different scenarios by learning users' cross-scenario interests via two specific attention modules, which leverage the scenario features and item features to modulate the user behavior features, respectively. Then, taking the encoded features of previous module as input, a scenario-specific linear transformation layer is adopted to further extract scenario-specific features, followed by two groups of debias expert networks, i.e., scenario-specific experts and scenario-shared experts. They output intermediate results independently, which are further fused into the final result by a multi-scenario gating module. In addition, to mitigate the data fairness issue caused by manual intervention, we propose the concept of Fairness Coefficient (FC) to measures the importance of individual sample and use it to reweigh the prediction in the debias expert networks. Experiments on an offline dataset covering over 80 million users and 1.55 million travel items and an online A/B test demonstrate the effectiveness of our SAR-Net and its superiority over state-of-the-art methods. SAR-Net has also been deployed in the online travel marketing platform of Alibaba and is serving hundreds of travel scenarios.

Conversion Rate (CVR) prediction in modern industrial e-commerce platforms is becoming increasingly important, which directly contributes to the final revenue. In order to address the well-known sample selection bias (SSB) and data sparsity (DS) issues encountered during CVR modeling, the abundant labeled macro behaviors (i.e., user's interactions with items) are used. Nonetheless, we observe that several purchase-related micro behaviors (i.e., user's interactions with specific components on the item detail page) can supplement fine-grained cues for CVR prediction. Motivated by this observation, we propose a novel CVR prediction method by Hierarchically Modeling both Micro and Macro behaviors (HM3). Specifically, we first construct a complete user sequential behavior graph to hierarchically represent micro behaviors and macro behaviors as one-hop and two-hop post-click nodes. Then, we embody HM3 as a multi-head deep neural network, which predicts six probability variables corresponding to explicit sub-paths in the graph. They are further combined into the prediction targets of four auxiliary tasks as well as the final CVR according to the conditional probability rule defined on the graph. By employing multi-task learning and leveraging the abundant supervisory labels from micro and macro behaviors, HM3 can be trained end-to-end and address the SSB and DS issues. Extensive experiments on both offline and online settings demonstrate the superiority of the proposed HM3 over representative state-of-the-art methods.

Matching items for a user from a travel item pool of large cardinality have been the most important technology for increasing the business at Fliggy, one of the most popular online travel platforms (OTPs) in China. There are three major challenges facing OTPs: sparsity, diversity, and implicitness. In this paper, we present a novel Fliggy ITinerary-aware deep matching NETwork (FitNET) to address these three challenges. FitNET is designed based on the popular deep matching network, which has been successfully employed in many industrial recommendation systems, due to its effectiveness. The concept itinerary is firstly proposed under the context of recommendation systems for OTPs, which is defined as the list of unconsumed orders of a user. All orders in a user itinerary are learned as a whole, based on which the implicit travel intention of each user can be more accurately inferred. To alleviate the sparsity problem, users’ profiles are incorporated into FitNET. Meanwhile, a series of itinerary-aware attention mechanisms that capture the vital interactions between user’s itinerary and other input categories are carefully designed. These mechanisms are very helpful in inferring a user’s travel intention or preference, and handling the diversity in a user’s need. Further, two training objectives, i.e., prediction accuracy of user’s travel intention and prediction accuracy of user’s click behavior, are utilized by FitNET, so that these two objectives can be optimized simultaneously. An offline experiment on Fliggy production dataset with over 0.27 million users and 1.55 million travel items, and an online A/B test both show that FitNET effectively learns users’ travel intentions, preferences, and diverse needs, based on their itineraries and gains superior performance compared with state-of-the-art methods. FitNET now has been successfully deployed at Fliggy, serving major online traffic.