Aili Chen, Chengyu Du, Jiangjie Chen, Jinghan Xu, Yikai Zhang, Siyu Yuan, Zulong Chen, Liangyue Li, Yanghua Xiao. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent advancements in large language models (LLMs) have demonstrated that progressive refinement, rather than providing a single answer, results in more accurate and thoughtful outputs. However, existing methods often rely heavily on supervision signals to evaluate previous responses, making it difficult to effectively assess output quality in more open-ended scenarios. Additionally, these methods are typically designed for specific tasks, which limits their generalization to new domains. To address these limitations, we propose Progressive Thought Refinement (PTR), a framework that enables LLMs to progressively refine their responses. PTR operates in two phases: (1) Thought data construction stage: We propose a weak and strong model collaborative selection strategy to build a high-quality progressive refinement dataset to ensure logical consistency from thought to answers, and the answers are gradually refined in each round. (2) Thought-Mask Fine-Tuning Phase: We design a training structure to mask the "thought" and adjust loss weights to encourage LLMs to refine prior thought, teaching them to implicitly understand "how to improve" rather than "what is correct." Experimental results show that PTR significantly enhances LLM performance across ten diverse tasks (avg. from 49.6% to 53.5%) without task-specific fine-tuning. Notably, in more open-ended tasks, LLMs also demonstrate substantial improvements in the quality of responses beyond mere accuracy, suggesting that PTR truly teaches LLMs to self-improve over time. Our work is now open-source. https://github.com/cydu24/Progressive-Thought-Refinement

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.

Abstract Hierarchical text classification (HTC) is an important task with broad applications, and few-shot HTC has gained increasing interest recently. While in-context learning (ICL) with large language models (LLMs) has achieved significant success in few-shot learning, it is not as effective for HTC because of the expansive hierarchical label sets and extremely ambiguous labels. In this work, we introduce the first ICL-based framework with LLM for few-shot HTC. We exploit a retrieval database to identify relevant demonstrations, and an iterative policy to manage multi-layer hierarchical labels. Particularly, we equip the retrieval database with HTC label-aware representations for the input texts, which is achieved by continual training on a pretrained language model with masked language modeling (MLM), layer-wise classification (CLS, specifically for HTC), and a novel divergent contrastive learning (DCL, mainly for adjacent semantically similar labels) objective. Experimental results on three benchmark datasets demonstrate superior performance of our method, and we can achieve state-of-the-art results in few-shot HTC.

Online Travel Platforms (OTPs) have been working on improving their hotel Search & Ranking (S&R) systems that facilitate efficient matching between consumers and hotels. Existing OTPs focus on improving platform revenue. In this work, we take a first step in incorporating hotel merchants' objectives into the design of hotel S&R systems to achieve an incentive loop: the OTP tilts impressions and better-ranked positions to merchants with high service quality, and in return, the merchants provide better service to consumers. Three critical design challenges need to be resolved to achieve this incentive loop: Matthew Effect in the consumer feedback-loop, unclear relation between hotel service quality and performance, and conflicts between platform revenue and consumer experience.

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.

Multi-sourced networks naturally appear in many application domains, ranging from bioinformatics, social networks, neuroscience to management. Although state-of-the-art offers rich models and algorithms to find various patterns when input networks are given, it has largely remained nascent on how vulnerable the mining results are due to the adversarial attacks. In this paper, we address the problem of attacking multi-network mining through the way of deliberately perturbing the networks to alter the mining results. The key idea of the proposed method admiring is effective and efficient influence functions on the Sylvester equation defined over the input networks, which plays a central and unifying role in various multi-network mining tasks. The proposed algorithms bear three main advantages, including (1) effectiveness, being able to accurately quantify the rate of change of the mining results in response to attacks; (2) efficiency, scaling linearly with more than 100 times speed-up over the straight-forward implementation without any quality loss; and (3) generality, being applicable to a variety of multi-network mining tasks (e.g., graph kernel, network alignment, cross-network node similarity) with different attacking strategies (e.g., edge/node removal, attribute alteration).

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.

Online travel platforms (OTPs), e.g., bookings.com and Ctrip.com, deliver travel experiences to online users by providing travel-related products. Although much progress has been made, the state-of-the-arts for cold-start problems are largely sub-optimal for user representation, since they do not take into account the unique characteristics exhibited from user travel behaviors. In this work, we propose a State-aware Multi-aspect Interests representation Network (SMINet) for cold-start users recommendation at OTPs, which consists of a multi-aspect interests extractor, a co-attention layer, and a state-aware gating layer. The key component of the model is the multi-aspect interests extractor, which is able to extract representations for the user's multi-aspect interests. Furthermore, to learn the interactions between the user behaviors in the current session and the above multi-aspect interests, we carefully design a co-attention layer which allows the cross attentions between the two modules. Additionally, we propose a travel state-aware gating layer to attentively select the multi-aspect interests. The final user representation is obtained by fusing the three components. Comprehensive experiments conducted both offline and online demonstrate the superior performance of the proposed model at user representation, especially for cold-start users, compared with state-of-the-art methods.

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.

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.

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.

Due to intense competition and lack of real estate on the front page of large e-commerce platforms, sellers are sometimes motivated to garner non-genuine signals (clicks, add-to-carts, purchases) on their products, to make them appear more appealing to customers. This hurts customers' trust on the platform, and also hurts genuine sellers who sell their items without looking to game the system. While it is important to find the sellers and the buyers who are colluding to garner these non-genuine signals, doing so is highly nontrivial. Firstly, the set of bad actors in the system is a very small fraction of all the buyers/sellers on the platform. Secondly, bad actors ``hide" with the good ones, making them hard to detect. In this paper, we develop CONGCN, a context aware heterogeneous graph convolutional network to detect bad actors on a large heterogeneous graph. While our method is motivated by abuse detection in e-commerce, the method is applicable to other areas such as computational biology and finance, where large heterogeneous graphs are pervasive, and the amount of labeled data is very limited. We train CONGCN via novel sampling methods, and context aware message passing in a semi-supervised fashion to predict dishonest buyers and sellers in e-commerce. Extensive experiments show that our method is effective, beating several baselines; generalizable to an inductive setting and highly scalable

Network alignment finds node correspondences across multiple networks, where the alignment accuracy is of crucial importance because of its profound impact on downstream applications. The vast majority of existing works focus on how to best utilize the topology and attribute information of the input networks as well as the anchor links when available. Nonetheless, it has not been well studied on how to boost the alignment performance through actively obtaining high-quality and informative anchor links, with a few exceptions. The sparse literature on active network alignment introduces the human in the loop to label some seed node correspondence (i.e., anchor links), which are informative from the perspective of querying the most uncertain node given few potential matchings. However, the direct influence of the intrinsic network attribute information on the alignment results has largely remained unknown. In this paper, we tackle this challenge and propose an active network alignment method (Attent) to identify the best nodes to query. The key idea of the proposed method is to leverage effective and efficient influence functions defined over the alignment solution to evaluate the goodness of the candidate nodes for query. Our proposed query strategy bears three distinct advantages, including (1) effectiveness, being able to accurately quantify the influence of the candidate nodes on the alignment results; (2) efficiency, scaling linearly with 15 − 17 × speed-up over the straight-forward implementation without any quality loss; (3) generality, consistently improving alignment performance of a variety of network alignment algorithms.

Teams can be often viewed as a dynamic system where the team configuration evolves over time (e.g., new members join the team; existing members leave the team; the skills of the members improve over time). Consequently, the performance of the team might be changing due to such team dynamics. A natural question is how to plan the (re-)staffing actions (e.g., recruiting a new team member) at each time step so as to maximize the expected cumulative performance of the team. In this paper, we address the problem of real-time team optimization by intelligently selecting the best candidates towards increasing the similarity between the current team and the high-performance teams according to the team configuration at each time-step. The key idea is to formulate it as a Markov Decision process (MDP) problem and leverage recent advances in reinforcement learning to optimize the team dynamically. The proposed method bears two main advantages, including (1) dynamics, being able to model the dynamics of the team to optimize the initial team towards the direction of a high-performance team via performance feedback; (2) efficacy, being able to handle the large state/action space via deep reinforcement learning based value estimation. We demonstrate the effectiveness of the proposed method through extensive empirical evaluations.

Multi-sourced networks naturally appear in many application domains, ranging from bioinformatics, social networks, neuroscience to management. Although state-of-the-art offers rich models and algorithms to find various patterns when input networks are given, it has largely remained nascent on how vulnerable the mining results are due to the adversarial attacks. In this paper, we address the problem of attacking multi-network mining through the way of deliberately perturbing the networks to alter the mining results. The key idea of the proposed method (Admiring) is effective influence functions on the Sylvester equation defined over the input networks, which plays a central and unifying role in various multi-network mining tasks. The proposed algorithms bear two main advantages, including (1) effectiveness, being able to accurately quantify the rate of change of the mining results in response to attacks; and (2) generality, being applicable to a variety of multi-network mining tasks ( e.g., graph kernel, network alignment, cross-network node similarity) with different attacking strategies (e.g., edge/node removal, attribute alteration).

Networked prediction has attracted lots of research attention in recent years. Compared with the traditional learning setting, networked prediction is even harder to understand due to its coupled, \em multi-level nature. The learning process propagates top-down through the underlying network from the macro level (the entire learning system), to meso level (learning tasks), and to micro level (individual learning examples). In the meanwhile, the networked prediction setting also offers rich context to explain the learning process through the lens of \em multi-aspect, including training examples ( e.g., what are the most influential examples ), the learning tasks ( e.g., which tasks are most important ) and the task network ( e.g., which task connections are the keys ). Thus, we propose a multi-aspect, multi-level approach to explain networked prediction. The key idea is to efficiently quantify the influence on different levels of the learning system due to the perturbation of various aspects. The proposed method offers two distinctive advantages: (1) multi-aspect, multi-level: it is able to explain networked prediction from multiple aspects (i.e., example-task-network) at multiple levels (i.e., macro-meso-micro); (2) efficiency: it has a linear complexity by efficiently evaluating the influences of changes to the networked prediction without retraining.

State-of-the-art in network science of teams offers effective recommendation methods to answer questions like who is the best replacement, what is the best team expansion strategy, but lacks intuitive ways to explain why the optimization algorithm gives the specific recommendation for a given team optimization scenario. To tackle this problem, we develop an interactive prototype system, Extra, as the first step towards addressing such a sense-making challenge, through the lens of the underlying network where teams embed, to explain the team recommendation results. The main advantages are (1) Algorithm efficacy: we propose an effective and fast algorithm to explain random walk graph kernel, the central technique for networked team recommendation; (2) Intuitive visual explanation: we present intuitive visual analysis of the recommendation results, which can help users better understand the rationality of the underlying team recommendation algorithm.

Teams are increasingly indispensable to achievements in any organization. Despite the organizations' substantial dependency on teams, fundamental knowledge about the conduct of team-enabled operations is lacking, especially at the social, cognitive and information level in relation to team performance and network dynamics.

Research impact plays a critical role in evaluating the research quality and influence of a scholar, a journal, or a conference. Many researchers have attempted to quantify research impact by introducing different types of metrics based on citation data, such as h -index, citation count, and impact factor. These metrics are widely used in the academic community. However, quantitative metrics are highly aggregated in most cases and sometimes biased, which probably results in the loss of impact details that are important for comprehensively understanding research impact. For example, which research area does a researcher have great research impact on? How does the research impact change over time? How do the collaborators take effect on the research impact of an individual? Simple quantitative metrics can hardly help answer such kind of questions, since more detailed exploration of the citation data is needed. Previous work on visualizing citation data usually only shows limited aspects of research impact and may suffer from other problems including visual clutter and scalability issues. To fill this gap, we propose an interactive visualization tool, ImpactVis , for better exploration of research impact through citation data. Case studies and in-depth expert interviews are conducted to demonstrate the effectiveness of ImpactVis .

The PART-WHOLE relationship routinely finds itself in many disciplines, ranging from collaborative teams, crowdsourcing, autonomous systems to networked systems. From the algorithmic perspective, the existing work has primarily focused on predicting the outcomes of the whole and parts, by either separate models or linear joint models, which assume the outcome of the parts has a linear and independent effect on the outcome of the whole. In this paper, we propose a joint predictive method named PAROLE to simultaneously and mutually predict the part and whole outcomes. The proposed method offers two distinct advantages over the existing work. First (Model Generality), we formulate joint PART-WHOLE outcome prediction as a generic optimization problem, which is able to encode a variety of complex relationships between the outcome of the whole and parts, beyond the linear independence assumption. Second (Algorithm Efficacy), we propose an effective and efficient block coordinate descent algorithm, which is able to find the coordinate-wise optimum with a linear complexity in both time and space. Extensive empirical evaluations on real-world datasets demonstrate that the proposed PAROLE (1) leads to consistent prediction performance improvement by modeling the non-linear part-whole relationship as well as part-part interdependency, and (2) scales linearly in terms of the size of the training dataset.

In this paper, we study ways to enhance the composition of teams based on new requirements in a collaborative environment. We focus on recommending team members who can maintain the team's performance by minimizing changes to the team's skills and social structure. Our recommendations are based on computing team-level similarity, which includes skill similarity, structural similarity as well as the synergy between the two. Current heuristic approaches are one-dimensional and not comprehensive, as they consider the two aspects independently. To formalize team-level similarity, we adopt the notion of graph kernel of attributed graphs to encompass the two aspects and their interaction. To tackle the computational challenges, we propose a family of fast algorithms by (a) designing effective pruning strategies, and (b) exploring the smoothness between the existing and the new team structures. Extensive empirical evaluations on real world datasets validate the effectiveness and efficiency of our algorithms.

The science of team science is a rapidly emerging research field that studies strategies to understand and enhance the process and outcomes of collaborative, team-based research. An interesting research question we address in this work is how to maintain and optimize the team performance should certain changes happen to the team. In particular, we take the network approach to understanding the teams and consider optimizing the teams with several operations (e.g., replacement, expansion, shrinkage). We develop TEAMOPT, a system to assist users in optimizing the team performance interactively to support the changes to a team. TEAMOPT takes as input a large network of individuals (e.g., co-author network of researchers) and is able to assist users in assembling a team with specific requirements and optimizing the team in response to the changes made to the team. It is effective in finding the best candidates, and interactive with users' feedback in the loop. The system is developed using HTML5, JavaScript, D3.js (front-end) and Python CGI (back-end). A prototype system is already deployed. We will invite the audience to experiment with our TEAMOPT in terms of its effectiveness, efficiency and applicability to various scenarios.

Measuring node proximity on large scale networks is a fundamental building block in many application domains, ranging from computer vision, e-commerce, social networks, software engineering, disaster management to biology and epidemiology. The state of the art (e.g., random walk based methods) typically assumes the input network is given a priori, with the known network topology and the associated edge weights. A few recent works aim to further infer the optimal edge weights based on the side information. This paper generalizes the challenge in multiple dimensions, aiming to learn optimal networks for node proximity measures. First (optimization scope), our proposed formulation explores a much larger parameter space, so that it is able to simultaneously infer the optimal network topology and the associated edge weights. This is important as a noisy or missing edge could greatly mislead the network node proximity measures. Second (optimization granularity), while all the existing works assume one common optimal network, be it given as the input or learned by the algorithms, exists for all queries, our method performs optimization at a much finer granularity, essentially being able to infer an optimal network that is specific to a given query. Third (optimization efficiency), we carefully design our algorithms with a linear complexity wrt the neighborhood size of the user preference set. We perform extensive empirical evaluations on a diverse set of 10+ real networks, which show that the proposed algorithms (1) consistently outperform the existing methods on all six commonly used metrics; (2) empirically scale sub-linearly to billion-scale networks and (3) respond in a fraction of a second.

Forecasting the success of scientific work has been attracting extensive research attention in the recent years. It is often of key importance to foresee the pathway to impact for scholarly entities for (1) tracking research frontier, (2) invoking an early intervention and (3) proactively allocating research resources. Many recent progresses have been seen in modeling the long-term scientific impact for point prediction. However, challenges still remain when it comes to forecasting the impact pathway. In this paper, we propose a novel predictive model to collectively achieve a set of design objectives to address these challenges, including prediction consistency and parameter smoothness. Extensive empirical evaluations on real scholarly data validate the effectiveness of the proposed model.

With the rapid growth of rich network data available through various sources such as social media and digital archives,there is a growing interest in more powerful network visual analysis tools and methods. The rich information about the network nodes and links can be represented as multivariate graphs, in which the nodes are accompanied with attributes to represent the properties of individual nodes. An important task often encountered in multivariate network analysis is to uncover link structure with groups, e.g., to understand why a person fits a specific job or certain role in a social group well.The task usually involves complex considerations including specific requirement of node attributes and link structure, and hence a fully automatic solution is typically not satisfactory.In this work, we identify the design challenges for min-ing groups with complex criteria and present an interactive system, "g-Miner," that enables visual mining of groups on multivariate graph data. We demonstrate the effectiveness of our system through case study and in-depth expert inter-views. This work contributes to understanding the design of systems for leveraging users' knowledge progressively with algorithmic capacity for tackling massive heterogeneous information.

Understanding the dynamic mechanisms that drive the high-impact scientific work (e.g., research papers, patents) is a long-debated research topic and has many important implications, ranging from personal career development and recruitment search, to the jurisdiction of research resources. Recent advances in characterizing and modeling scientific success have made it possible to forecast the long-term impact of scientific work, where data mining techniques, supervised learning in particular, play an essential role. Despite much progress, several key algorithmic challenges in relation to predicting long-term scientific impact have largely remained open. In this paper, we propose a joint predictive model to forecast the long-term scientific impact at the early stage, which simultaneously addresses a number of these open challenges, including the scholarly feature design, the non-linearity, the domain-heterogeneity and dynamics. In particular, we formulate it as a regularized optimization problem and propose effective and scalable algorithms to solve it. We perform extensive empirical evaluations on large, real scholarly data sets to validate the effectiveness and the efficiency of our method.

Graph kernels provide an expressive approach to measuring the similarity of two graphs, and are key building blocks behind many real-world applications, such as bioinformatics, brain science and social networks. However, current methods for computing graph kernels assume the input graphs are static, which is often not the case in reality. It is highly desirable to track the graph kernels on dynamic graphs evolving over time in a timely manner. In this paper, we propose a family of Cheetah algorithms to deal with the challenge. Cheetah leverages the low rank structure of graph updates and incrementally updates the eigen-decomposition or SVD of the adjacency matrices of graphs. Experimental evaluations on real world graphs validate our algorithms (1) are significantly faster than alternatives with high accuracy and (b) scale sub-linearly.

In this paper, we study the problem of TEAM MEMBER REPLACEMENT -- given a team of people embedded in a social network working on the same task, find a good candidate to best replace a team member who becomes unavailable to perform the task for certain reason (e.g., conflicts of interests or resource capacity). Prior studies in teamwork have suggested that a good team member replacement should bring synergy to the team in terms of having both skill matching and structure matching. However, existing techniques either do not cover both aspects or consider the two aspects independently. In this work, we propose a novel problem formulation using the concept of graph kernels that takes into account the interaction of both skill and structure matching requirements. To tackle the computational challenges, we propose a family of fast algorithms by (a) designing effective pruning strategies, and (b) exploring the smoothness between the existing and the new team structures. We conduct extensive experimental evaluations and user studies on real world datasets to demonstrate the effectiveness and efficiency. Our algorithms (a) perform significantly better than the alternative choices in terms of both precision and recall and (b) scale sub-linearly.

In this paper, we investigate the problem of recognizing occupations of multiple people with arbitrary poses in a photo. Previous work utilizing single person's nearly frontal clothing information and fore/background context preliminarily proves that occupation recognition is computationally feasible in computer vision. However, in practice, multiple people with arbitrary poses are common in a photo, and recognizing their occupations is even more challenging. We argue that with appropriately built visual attributes, co-occurrence, and spatial configuration model that is learned through structure SVM, we can recognize multiple people's occupations in a photo simultaneously. To evaluate our method's performance, we conduct extensive experiments on a new well-labeled occupation database with 14 representative occupations and over 7K images. Results on this database validate our method's effectiveness and show that occupation recognition is solvable in a more general case.