Offline reinforcement learning (RL) aims to learn policies without online explorations. To enlarge the training data, model-based offline RL learns a dynamics model which is utilized as a virtual environment to generate simulation data and enhance policy learning. However, existing data augmentation methods for offline RL suffer from (i) trivial improvement from short-horizon simulation; and (ii) the lack of evaluation and correction for generated data, leading to low-qualified augmentation.

ID-based Recommender Systems (RecSys), where each item is assigned a unique identifier and subsequently converted into an embedding vector, have dominated the de-signing of RecSys. Though prevalent, such ID-based paradigm is not suitable for developing transferable RecSys and is also susceptible to the cold -start issue. In this paper, we unleash the boundaries of the ID- based paradigm and propose a Pure Multi-Modality based Recommender system (PMMRec), which relies solely on the multi-modal contents of the items (e.g., texts and images) and learns transition patterns general enough to transfer across domains and platforms. Specifically, we design a plug-and-play framework architecture consisting of multi-modal item encoders, a fusion module, and a user encoder. To align the cross-modal item representations, we propose a novel next-item enhanced cross-modal contrastive learning objective, which is equipped with both inter- and intra-modality negative samples and explicitly incorporates the transition patterns of user behaviors into the item encoders. To ensure the robustness of user representations, we propose a novel noised item detection objective and a robustness-aware contrastive learning objective, which work together to denoise user sequences in a self-supervised manner. PMMRec is designed to be loosely coupled, so after being pre-trained on the source data, each component can be transferred alone, or in conjunction with other components, allowing PMMRec to achieve versatility under both multi-modality and single-modality transfer learning settings. Extensive experiments on 4 sources and 10 target datasets demonstrate that PMMRec surpasses the state-of-the-art recommenders in both recommendation performance and transferability. Our code and dataset is available at: https://github.com/ICDE24IPMMRec.

The advent of large language models has inspired active and promising research focused on developing text content-based recommendation models. Meanwhile, although image features are also key signals in recommender systems, there is currently a lack of research on recommendation models that are primarily based on raw image pixels. The lack of large-scale datasets containing raw images in visually driven recommendation scenarios has been a significant barrier to the development of this research direction. To address this challenge, we introduce PixelRec, a comprehensive dataset of cover images collected from a video streaming platform. With approximately 200 million user image interactions, 30 million users, and 400,000 high-resolution short video cover images, PixelRec facilitates the development, benchmarking, and analysis of various image pixel based recommendation models. Leveraging this dataset, we establish a accessible pipeline to implement a series of vision-based recommendation models, providing extensive benchmark results for them. Our contributions include the PixelRec dataset, baseline algorithms, operational pipeline, exploratory findings, and the PixelRec benchmark. We believe PixelRec will significantly advance research on recommendation models based on image content and foster fruitful collaboration between the fields of recommender systems and computer vision. The dataset, code, and documents are made available at https://github.com/westlake-repl/PixelRec.

With the rise of large language models (LLMs), recent works have leveraged LLMs to improve the performance of click-through rate (CTR) prediction. However, we argue that a critical obstacle remains in deploying LLMs for practical use: the efficiency of LLMs when processing long textual user behaviors. As user sequences grow longer, the current efficiency of LLMs is inadequate for training on billions of users and items. To break through the efficiency barrier of LLMs, we propose Behavior Aggregated Hierarchical Encoding (BAHE) to enhance the efficiency of LLM-based CTR modeling. Specifically, BAHE proposes a novel hierarchical architecture that decouples the encoding of user behaviors from inter-behavior interactions. Firstly, to prevent computational redundancy from repeated encoding of identical user behaviors, BAHE employs the LLM's pre-trained shallow layers to extract embeddings of the most granular, atomic user behaviors from extensive user sequences and stores them in the offline database. Subsequently, the deeper, trainable layers of the LLM facilitate intricate inter-behavior interactions, thereby generating comprehensive user embeddings. This separation allows the learning of high-level user representations to be independent of low-level behavior encoding, significantly reducing computational complexity. Finally, these refined user embeddings, in conjunction with correspondingly processed item embeddings, are incorporated into the CTR model to compute the CTR scores. Extensive experimental results show that BAHE reduces training time and memory by five times for CTR models using LLMs, especially with longer user sequences. BAHE has been deployed in a real-world system, allowing for daily updates of 50 million CTR data on 8 A100 GPUs, making LLMs practical for industrial CTR prediction.

Click-through rate (CTR) prediction plays a crucial role in recommendation systems, with significant impact on user experience and platform revenue generation. Despite the various public CTR datasets available due to increasing interest from both academia and industry, these datasets have limitations. They cover a limited range of scenarios and predominantly focus on ID-based features, neglecting the vital role of multi-modal features for effective multi-scenario CTR prediction. Moreover, their scale is modest compared to real-world industrial datasets, hindering robust and comprehensive evaluation of complex models. To address these challenges, we introduce a large-scale Multi-Scenario Multi-Modal CTR dataset named AntM2 C, built from real industrial data from Alipay. This dataset offers an impressive breadth and depth of information, covering CTR data from four diverse business scenarios, including advertisements, consumer coupons, mini-programs, and videos. Unlike existing datasets, AntM2 C provides not only ID-based features but also five textual features and one image feature for both users and items, supporting more delicate multi-modal CTR prediction. AntM2 C is also substantially larger than existing datasets, comprising 100 million CTR data. This scale allows for robust and comprehensive evaluation and comparison of CTR prediction models. We employ AntM2 C to construct several typical CTR tasks, including multi-scenario modeling, item and user cold-start modeling, and multi-modal modeling. Initial experiments and comparisons with baseline methods have shown that AntM2 C presents both new challenges and opportunities for CTR models, with the potential to significantly advance CTR research. The AntM2 C dataset is available at https://www.atecup.cn/OfficalDataSet.

Adapters, a plug-in neural network module with some tunable parameters, have emerged as a parameter-efficient transfer learning technique for adapting pre-trained models to downstream tasks, especially for natural language processing (NLP) and computer vision (CV) fields. Meanwhile, learning recommendation models directly from raw item modality features --- e.g., texts of NLP and images of CV --- can enable effective and transferable recommender systems (called TransRec). In view of this, a natural question arises:can adapter-based learning techniques achieve parameter-efficient TransRec with good performance?

Recommendation models that utilize unique identities (IDs for short) to represent distinct users and items have been state-of-the-art (SOTA) and dominated the recommender systems (RS) literature for over a decade. Meanwhile, the pre-trained modality encoders, such as BERT [9] and Vision Transformer [11], have become increasingly powerful in modeling the raw modality features of an item, such as text and images. Given this, a natural question arises: can a purely modality-based recommendation model (MoRec) outperforms or matches a pure ID-based model (IDRec) by replacing the itemID embedding with a SOTA modality encoder? In fact, this question was answered ten years ago when IDRec beats MoRec by a strong margin in both recommendation accuracy and efficiency.

Making accurate recommendations for cold-start users has been a longstanding and critical challenge for recommender systems (RS). Cross-domain recommendations (CDR) offer a solution to tackle such a cold-start problem when there is no sufficient data for the users who have rarely used the system. An effective approach in CDR is to leverage the knowledge (e.g., user representations) learned from a related but different domain and transfer it to the target domain. Fine-tuning works as an effective transfer learning technique for this objective, which adapts the parameters of a pre-trained model from the source domain to the target domain. However, current methods are mainly based on the global fine-tuning strategy: the decision of which layers of the pre-trained model to freeze or fine-tune is taken for all users in the target domain. In this paper, we argue that users in RS are personalized and should have their own fine-tuning policies for better preference transfer learning. As such, we propose a novel User-specific Adaptive Fine-tuning method (UAF), selecting which layers of the pre-trained network to fine-tune, on a per-user basis. Specifically, we devise a policy network with three alternative strategies to automatically decide which layers to be fine-tuned and which layers to have their parameters frozen for each user. Extensive experiments show that the proposed UAF exhibits significantly better and more robust performance for user cold-start recommendation.

Recently, Memory-based Neural Recommenders (MNR) have demonstrated superior predictive accuracy in the task of sequential recommendations, particularly for modeling long-term item dependencies. However, typical MNR requires complex memory access operations, i.e., both writing and reading via a controller (e.g., RNN) at every time step. Those frequent operations will dramatically increase the network training time, resulting in the difficulty in being deployed on industrial-scale recommender systems. In this paper, we present a novel generalChunkframework to accelerate MNR significantly. Specifically, our framework divides proximal information units into chunks, and performs memory access at certain time steps, whereby the number of memory operations can be greatly reduced. We investigate two ways to implement effective chunking, i.e., PEriodic Chunk (PEC) and Time-Sensitive Chunk (TSC), to preserve and recover important recurrent signals in the sequence. Since chunk-accelerated MNR models take into account more proximal information units than that from a single timestep, it can alleviate the influence of noise in the user-item interaction sequence to a large extent, and thus improve the stability of MNR. In this way, the proposed chunk mechanism can lead to not only faster training and prediction, but even slightly better results. The experimental results on three real-world datasets (weishi, ml-10M and ml-latest) show that our chunk framework notably reduces the running time (e.g., with up to 7x for training & 10x for inference on ml-latest) of MNR, and meantime achieves competitive performance.

Deep neural networks (DNN) based recommender models often require numerous parameters to achieve remarkable performance. However, this inevitably brings redundant neurons, a phenomenon referred to as over-parameterization. In this paper, we plan to exploit such redundancy phenomena for recommender systems (RS), and propose a top-N item recommendation framework called PCRec that leverages collaborative training of two recommender models of the same network structure, termed peer collaboration. We first introduce two criteria to identify the importance of parameters of a given recommender model. Then, we rejuvenate the unimportant parameters by copying parameters from its peer network. After such an operation and retraining, the original recommender model is endowed with more representation capacity by possessing more functional model parameters. To show its generality, we instantiate PCRec by using three well-known recommender models. We conduct extensive experiments on two real-world datasets, and show that PCRec yields significantly better performance than its counterpart with the same model (parameter) size.

Learning effective user representations from sequential user-item interactions is a fundamental problem for recommender systems (RS). Recently, several unsupervised methods focusing on user representations pre-training have been explored. In general, these methods apply similar learning paradigms by first corrupting the behavior sequence, and then restoring the original input with some item-level prediction loss functions. Despite its effectiveness, we argue that there exist important gaps between such item-level optimization objective and user-level representations, and as a result, the learned user representations may only lead to sub-optimal generalization performance. In this paper, we propose a novel self-supervised pre-training framework, called CLUE, which stands for employing Contrastive Learning for modeling sequence-level User rEpresentation. The core idea of CLUE is to regard each user behavior sequence as a whole and then construct the self-supervision signals by transforming the original user behaviors by data augmentations (DA). Specifically, we employ two Siamese (weight-sharing) networks to learn the user-oriented representations, where the optimization goal is to maximize the similarity of learned representations of the same user by these two encoders. More importantly, we perform careful investigation of the impacts of view generating strategies for user behavior inputs from a more comprehensive perspective, including processing sequential behaviors by explicit DA strategies and employing dropout as implicit DA. To verify the effectiveness of CLUE, we perform extensive experiments on several user-related tasks with different scales and characteristics. Our experimental results show that the user representations learned by CLUE surpass existing item-level baselines under several evaluation protocols.

One-hot encoder accompanied by a softmax loss has become the default configuration to deal with the multiclass problem, and is also prevalent in deep learning (DL) based recommender systems (RS). The standard learning process of such methods is to fit the model outputs to a one-hot encoding of the ground truth, referred to as the hard target. However, it is known that these hard targets largely ignore the ambiguity of unobserved feedback in RS, and thus may lead to sub-optimal generalization performance. In this work, we propose SoftRec, a new RS optimization framework to enhance item recommendation. The core idea is that we add additional supervisory signals - well-designed soft targets - for each instance so as to better guide the recommender learning. Meanwhile, we carefully investigate the impacts of specific soft target distributions by instantiating the SoftRec with a series of strategies, including item-based, user-based, and model-based. To verify the effectiveness of SoftRec, we conduct extensive experiments on two public recommendation datasets by using various deep recommendation architectures. The experimental results show that our methods achieve superior performance compared with the standard optimization approaches. Moreover, SoftRec could also exhibit strong performance in cold-start scenarios where user-item interaction has higher sparsity.

Lifelong learning capabilities are crucial for sentiment classifiers to process continuous streams of opinioned information on the Web. However, performing lifelong learning is non-trivial for deep neural networks as continually training of incrementally available information inevitably results in catastrophic forgetting or interference. In this paper, we propose a novel i terative network p runing with uncertainty r egularization method for l ifelong s entiment classification (IPRLS), which leverages the principles of network pruning and weight regularization. By performing network pruning with uncertainty regularization in an iterative manner, IPRLS can adapt a single BERT model to work with continuously arriving data from multiple domains while avoiding catastrophic forgetting and interference. Specifically, we leverage an iterative pruning method to remove redundant parameters in large deep networks so that the freed-up space can then be employed to learn new tasks, tackling the catastrophic forgetting problem. Instead of keeping the old-tasks fixed when learning new tasks, we also use an uncertainty regularization based on the Bayesian online learning framework to constrain the update of old tasks weights in BERT, which enables positive backward transfer, i.e. learning new tasks improves performance on past tasks while protecting old knowledge from being lost. In addition, we propose a task-specific low-dimensional residual function in parallel to each layer of BERT, which makes IPRLS less prone to losing the knowledge saved in the base BERT network when learning a new task. Extensive experiments on 16 popular review corpora demonstrate that the proposed IPRLS method significantly outperforms the strong baselines for lifelong sentiment classification. For reproducibility, we submit the code and data at: \urlhttps://github.com/siat-nlp/IPRLS .

Deep learning has brought great progress for the sequential recommendation (SR) tasks. With advanced network architectures, sequential recommender models can be stacked with many hidden layers, e.g., up to 100 layers on real-world recommendation datasets. Training such a deep network is difficult because it can be computationally very expensive and takes much longer time, especially in situations where there are tens of billions of user-item interactions. To deal with such a challenge, we present StackRec, a simple, yet very effective and efficient training framework for deep SR models by iterative layer stacking. Specifically, we first offer an important insight that hidden layers/blocks in a well-trained deep SR model have very similar distributions. Enlightened by this, we propose the stacking operation on the pre-trained layers/blocks to transfer knowledge from a shallower model to a deep model, then we perform iterative stacking so as to yield a much deeper but easier-to-train SR model. We validate the performance of StackRec by instantiating it with four state-of-the-art SR models in three practical scenarios with real-world datasets. Extensive experiments show that StackRec achieves not only comparable performance, but also substantial acceleration in training time, compared to SR models that are trained from scratch. Codes are available at https://github.com/wangjiachun0426/StackRec.

Learning user representations is a vital technique toward effective user modeling and personalized recommender systems. Existing approaches often derive an individual set of model parameters for each task by training on separate data. However, the representation of the same user potentially has some commonalities, such as preference and personality, even in different tasks. As such, these separately trained representations could be suboptimal in performance as well as inefficient in terms of parameter sharing. In this paper, we delve on research to continually learn user representations task by task, whereby new tasks are learned while using partial parameters from old ones. A new problem arises since when new tasks are trained, previously learned parameters are very likely to be modified, and as a result, an artificial neural network (ANN)-based model may lose its capacity to serve for well-trained previous tasks forever, this issue is termed catastrophic forgetting. To address this issue, we present Conure the first continual, or lifelong, user representation learner --- i.e., learning new tasks over time without forgetting old ones. Specifically, we propose iteratively removing less important weights of old tasks in a deep user representation model, motivated by the fact that neural network models are usually over-parameterized. In this way, we could learn many tasks with a single model by reusing the important weights, and modifying the less important weights to adapt to new tasks. We conduct extensive experiments on two real-world datasets with nine tasks and show that Conure largely exceeds the standard model that does not purposely preserve such old "knowledge'', and performs competitively or sometimes better than models which are trained either individually for each task or simultaneously by merging all task data.

Sequential recommender systems (SRS) have become the key technology in capturing user's dynamic interests and generating high-quality recommendations. Current state-of-the-art sequential recommender models are typically based on a sandwich-structured deep neural network, where one or more middle (hidden) layers are placed between the input embedding layer and output softmax layer. In general, these models require a large number of parameters to obtain optimal performance. Despite the effectiveness, at some point, further increasing model size may be harder for model deployment in resource-constraint devices. To resolve the issues, we propose a compressed sequential recommendation framework, termed as CpRec, where two generic model shrinking techniques are employed. Specifically, we first propose a block-wise adaptive decomposition to approximate the input and softmax matrices by exploiting the fact that items in SRS obey a long-tailed distribution. To reduce the parameters of the middle layers, we introduce three layer-wise parameter sharing schemes. We instantiate CpRec using deep convolutional neural network with dilated kernels given consideration to both recommendation accuracy and efficiency. By the extensive ablation studies, we demonstrate that the proposed CpRec can achieve up to 4~8 times compression rates in real-world SRS datasets. Meanwhile, CpRec is faster during training & inference, and in most cases outperforms its uncompressed counterpart.

Inductive transfer learning has had a big impact on computer vision and NLP domains but has not been used in the area of recommender systems. Even though there has been a large body of research on generating recommendations based on modeling user-item interaction sequences, few of them attempt to represent and transfer these models for serving downstream tasks where only limited data exists.

Session-based recommender systems have attracted much attention recently. To capture the sequential dependencies, existing methods resort either to data augmentation techniques or left-to-right style autoregressive training. Since these methods are aimed to model the sequential nature of user behaviors, they ignore the future data of a target interaction when constructing the prediction model for it. However, we argue that the future interactions after a target interaction, which are also available during training, provide valuable signal on user preference and can be used to enhance the recommendation quality.

With the prevalence of multimedia content on the Web, developing recommender solutions that can effectively leverage the rich signal in multimedia data is in urgent need. Owing to the success of deep neural networks in representation learning, recent advances on multimedia recommendation has largely focused on exploring deep learning methods to improve the recommendation accuracy. To date, however, there has been little effort to investigate the robustness of multimedia representation and its impact on the performance of multimedia recommendation. In this paper, we shed light on the robustness of multimedia recommender system. Using the state-of-the-art recommendation framework and deep image features, we demonstrate that the overall system is not robust, such that a small (but purposeful) perturbation on the input image will severely decrease the recommendation accuracy. This implies the possible weakness of multimedia recommender system in predicting user preference, and more importantly, the potential of improvement by enhancing its robustness. To this end, we propose a novel solution named Adversarial Multimedia Recommendation (AMR), which can lead to a more robust multimedia recommender model by using adversarial learning. The idea is to train the model to defend an adversary, which adds perturbations to the target image with the purpose of decreasing the model's accuracy. We conduct experiments on two representative multimedia recommendation tasks, namely, image recommendation and visually-aware product recommendation. Extensive results verify the positive effect of adversarial learning and demonstrate the effectiveness of our AMR method. Source codes are available in https://github.com/duxy-me/AMR.

Convolutional Neural Networks (CNNs) have been recently introduced in the domain of session-based next item recommendation. An ordered collection of past items the user has interacted with in a session (or sequence) are embedded into a 2-dimensional latent matrix, and treated as an image. The convolution and pooling operations are then applied to the mapped item embeddings. In this paper, we first examine the typical session-based CNN recommender and show that both the generative model and network architecture are suboptimal when modeling long-range dependencies in the item sequence. To address the issues, we introduce a simple, but very effective generative model that is capable of learning high-level representation from both short- and long-range item dependencies. The network architecture of the proposed model is formed of a stack of holed convolutional layers, which can efficiently increase the receptive fields without relying on the pooling operation. Another contribution is the effective use of residual block structure in recommender systems, which can ease the optimization for much deeper networks. The proposed generative model attains state-of-the-art accuracy with less training time in the next item recommendation task. It accordingly can be used as a powerful recommendation baseline to beat in future, especially when there are long sequences of user feedback.

As the core of recommender systems, collaborative filtering (CF) models the affinity between a user and an item from historical user-item interactions, such as clicks, purchases, and so on. Benefiting from the strong representation power, neural networks have recently revolutionized the recommendation research, setting up a new standard for CF. However, existing neural recommender models do not explicitly consider the correlations among embedding dimensions, making them less effective in modeling the interaction function between users and items. In this work, we emphasize on modeling the correlations among embedding dimensions in neural networks to pursue higher effectiveness for CF. We propose a novel and general neural collaborative filtering framework—namely, ConvNCF, which is featured with two designs: (1) applying outer product on user embedding and item embedding to explicitly model the pairwise correlations between embedding dimensions, and (2) employing convolutional neural network above the outer product to learn the high-order correlations among embedding dimensions. To justify our proposal, we present three instantiations of ConvNCF by using different inputs to represent a user and conduct experiments on two real-world datasets. Extensive results verify the utility of modeling embedding dimension correlations with ConvNCF, which outperforms several competitive CF methods.

Although the word-popularity based negative sampler has shown superb performance in the skip-gram model, the theoretical motivation behind oversampling popular (non-observed) words as negative samples is still not well understood. In this paper, we start from an investigation of the gradient vanishing issue in the skip-gram model without a proper negative sampler. By performing an insightful analysis from the stochastic gradient descent (SGD) learning perspective, we demonstrate, both theoretically and intuitively, negative samples with larger inner product scores are more informative than those with lower scores for the SGD learner in terms of both convergence rate and accuracy. Understanding this, we propose an alternative sampling algorithm that dynamically selects informative negative samples during each SGD update. More importantly, the proposed sampler accounts for multi-dimensional self-embedded features during the sampling process, which essentially makes it more effective than the original popularity-based (one-dimensional) sampler. Empirical experiments further verify our observations and show that our fine-grained samplers gain significant improvement over the existing ones without increasing computational complexity.

To cope with ambiguous and/or underspecified queries, search result diversification (SRD) is a key technique that has attracted a lot of attention. This paper focuses on implicit SRD, where the possible subtopics underlying a query are unknown beforehand. We formulate implicit SRD as a process of selecting and ranking k exemplar documents that utilizes integer linear programming (ILP). Unlike the common practice of relying on approximate methods, this formulation enables us to obtain the optimal solution of the objective function. Based on four benchmark collections, our extensive empirical experiments reveal that: (1) The factors, such as different initial runs, the number of input documents, query types and the ways of computing document similarity significantly affect the performance of diversification models. Careful examinations of these factors are highly recommended in the development of implicit SRD methods. (2) The proposed method can achieve substantially improved performance over the state-of-the-art unsupervised methods for implicit SRD.

In the age of Web 2.0, a substantial amount of unstructured content are distributed through multiple text streams in an asynchronous fashion, which makes it increasingly difficult to glean and distill useful information. An effective way to explore the information in text streams is topic modelling, which can further facilitate other applications such as search, information browsing, and pattern mining. In this paper, we propose a semantic graph based topic modelling approach for structuring asynchronous text streams. Our model integrates topic mining and time synchronization, two core modules for addressing the problem, into a unified model. Specifically, for handling the lexical gap issues, we use global semantic graphs of each timestamp for capturing the hidden interaction among entities from all the text streams. For dealing with the sources asynchronism problem, local semantic graphs are employed to discover similar topics of different entities that can be potentially separated by time gaps. Our experiment on two real-world datasets shows that the proposed model significantly outperforms the existing ones.

State-of-the-art item recommendation algorithms, which apply Factorization Machines (FM) as a scoring function and pairwise ranking loss as a trainer (PRFM for short), have been recently investigated for the implicit feedback based context-aware recommendation problem (IFCAR). However, good recommenders particularly emphasize on the accuracy near the top of the ranked list, and typical pairwise loss functions might not match well with such a requirement. In this paper, we demonstrate, both theoretically and empirically, PRFM models usually lead to non-optimal item recommendation results due to such a mismatch. Inspired by the success of LambdaRank, we introduce Lambda Factorization Machines (LambdaFM), which is particularly intended for optimizing ranking performance for IFCAR. We also point out that the original lambda function suffers from the issue of expensive computational complexity in such settings due to a large amount of unobserved feedback. Hence, instead of directly adopting the original lambda strategy, we create three effective lambda surrogates by conducting a theoretical analysis for lambda from the top-N optimization perspective. Further, we prove that the proposed lambda surrogates are generic and applicable to a large set of pairwise ranking loss functions. Experimental results demonstrate LambdaFM significantly outperforms state-of-the-art algorithms on three real-world datasets in terms of four standard ranking measures.

Community Question Answering (CQA) services, such as Yahoo! Answers and WikiAnswers, have become popular with users as one of the central paradigms for satisfying users' information needs. The task of question retrieval aims to resolve one's query directly by finding the most relevant questions (together with their answers) from an archive of past questions. However, as the text of each question is short, there is usually a lexical gap between the queried question and the past questions. To alleviate this problem, we present a hybrid approach that blends several language modelling techniques for question retrieval, namely, the classic (query-likelihood) language model, the state-of-the-art translation-based language model, and our proposed semantics-based language model. The semantics of each candidate question is given by a probabilistic topic model which makes use of local and global semantic graphs for capturing the hidden interactions among entities (e.g., people, places, and concepts) in question-answer pairs. Experiments on two real-world datasets show that our approach can significantly outperform existing ones.