In this work, we propose VerbaLightGCN, a novel LLM-based recommendation framework that integrates the semantic understanding of LLMs with user-item interaction modeling. Traditional collaborative filtering (CF) models typically embed user and item IDs into a latent space to capture interaction signals. However, pretrained LLMs cannot natively interpret these learned embeddings. To bridge this gap, VerbaLightGCN adopts a CF-as-text paradigm, in which collaborative signals are encoded in textual form and directly learned from the user–item interaction graph, and are then combined with semantic information to construct user and item profiles that function as latent embeddings. Inspired by LightGCN, our method retains its message-passing design but replaces numerical embedding computations with a Chain-of-Thought prompting mechanism. This enables LLMs to simulate the LightGCN aggregation process through natural language. The result is a recommendation framework that unifies semantic understanding with collaborative signals in a fully language-native form. Experiments show that VerbaLightGCN achieves superior performance to both zero-shot LLM-based and traditional CF-based baselines. Further analysis reveals that the user and item profiles generated by VerbaLightGCN effectively capture both semantic preferences and collaborative filtering signals.

Disentangled recommendation within the Variational Autoencoder (VAE) framework aims to capture multiple user interests. While effective, these VAEs are fundamentally constrained by their reliance on interaction data alone, lacking the rich external semantic knowledge needed to properly structure and separate latent interests. Meanwhile, Large Language Models (LLMs) excel at deriving profound user preference signals from textual data. Prevailing methods for integrating LLMs into recommendation, however, either focus on single-interest modeling or perform a shallow fusion by aligning LLM and VAE representation spaces. Thus, they fail to fundamentally shape the VAE's latent space for multi-interest learning, hindering recommendation performance. To bridge this gap, we propose to fundamentally shift the integration point: instead of aligning representation spaces, we bridge the LLM-generated embedding space directly to the VAE's parameter space. Our framework designs hypernetworks to generate parameters governing interest discovery of a disentangled VAE conditioned on LLM-derived user embeddings. This direct bridge injects rich semantic knowledge into the model's learning foundation, preserving the VAE's power for disentangled representation while dramatically enhancing its modeling capability with informative, LLM-structured priors. Extensive experiments on benchmark datasets show that our method yields significant performance gains, unveiling the untapped potential of using LLMs for disentangled recommendation.

In the online digital realm, recommendation systems are ubiquitous and play a crucial role in enhancing user experience. These systems leverage user preferences to provide personalized recommendations, thereby helping users navigate through the paradox of choice. This work focuses on personalized sequential recommendation, where the system considers not only a user’s immediate, evolving session context, but also their cumulative historical behavior to provide highly relevant and timely recommendations. Through an empirical study conducted on diverse real-world datasets, we have observed and quantified the existence and impact of both short-term (immediate and transient) and long-term (enduring and stable) preferences on users’ historical interactions. Building on these insights, we propose a framework that combines short- and long-term preferences to enhance recommendation performance, namely Compositions of Variant Experts ( CoVE ). This novel framework dynamically integrates short- and long-term preferences through the use of different specialized recommendation models (i.e., experts). Extensive experiments showcase the effectiveness of the proposed methods and ablation studies further investigate the impact of variant expert types.

Learning effective latent representations for users and items is the cornerstone of recommender systems. Traditional approaches rely on user-item interaction data to map users and items into a shared latent space, but the limiting factor is often data sparsity. While reviews could mitigate this sparsity, existing review-aware recommendation models may not have fully exploited their potential. First, they typically rely heavily on reviews as side information or additional features, but reviews are not universal, with many users and items lacking them. Second, such approaches use reviews as supplementary information, allowing potential divergence or inconsistency between the review representations and the user-item space. To overcome these limitations, our work introduces a Review-centric Contrastive Alignment Framework for Recommendation (ReCAFR), which incorporates reviews into the core learning process, ensuring alignment among user, item, and review representations within a unified space. Specifically, we leverage two self-supervised contrastive strategies that not only exploit the review augmentation to alleviate sparsity, but also align the tripartite representations to enhance robustness. Empirical studies on public benchmark datasets demonstrate the effectiveness and robustness of ReCAFR

Frameworks for discovering multiple user interest factors based on Variational AutoEncoder (VAE) has demonstrated competitive recommendation performance. However, as VAE only considers one user as input at a time, sharing across like-minded users may not be adequately facilitated. Moreover, interest sharing between users is not always available and thus, poses a challenge for VAE to explicitly model this information. To resolve this, we introduce an inter-user memory-based mechanism to unsupervisedly discover latent interest sharing between users under VAE framework. Concretely, we design a memory including an array of prototypes, each hypothetically representing a group of users sharing a particular interest. These memory prototypes are jointly trained with the backbone VAE-based recommendation model. For each user, we first discover multiple intra-user interest factors behind their item adoptions. Next, intra-user interest factors query to memory to retrieve the inter-user interest clues from like-minded users. This query-retrieve process is performed sequentially via a series of attention-transformation steps. Then, interest clues retrieved from memory are incorporated into interest factor representations of each user to increase their expressiveness. Thorough experiments on real-world datasets verify the strength of our method over an array of baselines. We further conduct qualitative analysis to understand the inner working of our memory-based refinement approach.

No abstract available.

Texts are often interconnected in a network structure, e.g., academic papers via citations. On the one hand, though Graph Neural Networks (GNNs) have shown promising ability to derive effective embeddings for networked documents, they do not assume latent topics, resulting in uninterpretahle embeddings. On the other hand, topic models can infer interpretable document representations. However, most topic models focus on plain text and fail to leverage network structure across documents. In this paper, we propose a GNN-based topic model that both captures network connection and derives semantically interpretable text representations. For network modeling, we build our model with Optimal Transport Barycenter. For semantic interpretability, we extend optimal transport with pre-trained word embeddings.

Recommendation explanations help to improve their acceptance by end users. Explanations come in many different forms. One that is of interest here is presenting an existing review of the recommended item as the explanation. The challenge is in selecting a suitable review, which is customarily addressed by assessing the relative importance or “attention” of each review to the recommendation objective. Our focus is improving review-level explanation by leveraging additional information in the form of questions and answers (QA). The proposed framework employs QA in an attention mechanism that aligns reviews to various QAs of an item and assesses their contribution jointly to the recommendation objective. The benefits are two-fold. For one, QA aids in selecting more useful reviews. For another, QA itself could accompany a well-aligned review in an expanded form of explanation. Experiments on datasets of 10 product categories showcase the efficacies of our method as compared to comparable baselines in identifying useful reviews and QAs, while maintaining parity in recommendation performance.

How can we adapt the composition of a post over a series of rounds to make it more appealing in a social network? Techniques that progressively learn how to make a fixed post more influential over rounds have been studied in the context of the Influence Maximization (IM) problem, which seeks a set of seed users that maximize a post’s influence. However, there is no work on progressively learning how a post’s features affect its influence. In this article, we propose and study the problem of Adaptive Content-Aware Influence Maximization (ACAIM), which calls to find k features to form a post in each round so as to maximize the cumulative influence of those posts over all rounds. We solve ACAIM by applying, for the first time, an Online Learning to Rank (OLR) framework for IM purposes. We introduce the CATRID propagation model , which expresses how posts disseminate in a social network using click probabilities and post visibility criteria and develop a simulator that runs CATRID via a training-testing scheme based on real posts of the VK social network, so as to realistically represent the learning environment. We deploy three learners that solve ACAIM in an online (real-time) manner. We experimentally prove the practical suitability of our solutions via exhaustive experiments on multiple brands (operating as different case studies ) and several VK datasets; the best learner is evaluated on 45 separate case studies yielding convincing results.

Text documents are often interconnected in a network structure, e.g., academic papers via citations, Web pages via hyperlinks. On the one hand, though Graph Neural Networks (GNNs) have shown promising ability to derive effective embeddings for such networked documents, they do not assume a latent topic structure and result inuninterpretableembeddings. On the other hand, topic models can infer semantically interpretable topic distributions for documents by associating each topic with a group of understandable key words. However, most topic models mainly focus on plain text within documents and fail to leveragenetwork structureacross documents. Network connectivity reveals topic similarity between linked documents, and modeling it could uncover meaningful semantics. Motivated by above two challenges, in this paper, we propose a GNN-based neural topic model that both captures network connectivity and derives semantically interpretable topic distributions for networked documents. For network modeling, we build the model based on the theory of Optimal Transport Barycenter, which captures network structure by allowing the topic distribution of a document to generate the content of its linked neighbors. For semantic interpretability, we extend optimal transport by incorporating semantically related words in the embedding space. Since Dirichlet prior in Latent Dirichlet Allocation successfully improves topic quality, we also analyze Dirichlet as an optimal transport prior distribution to improve topic interpretability. We design rejection sampling to simulate Dirichlet distribution. Extensive experiments on document classification, clustering, link prediction, and topic analysis verify the effectiveness of our model.

Variational Autoencoder (VAE) offers a non-linear probabilistic modeling of user's preferences. While it has achieved remarkable performance at collaborative filtering, it typically samples a single vector for representing user's preferences, which may be insufficient to capture the user's diverse interests. Existing solutions extend VAE to model multiple interests of users by resorting a variant of self-attentive method, i.e., employing prototypes to group items into clusters, each capturing one topic of user's interests. Despite showing improvements, the current design could be more effective since prototypes are randomly initialized and shared across users, resulting in uninformative and non-personalized clusters.

Connectivity across documents often exhibits a hierarchical network structure. Hyperbolic Graph Neural Networks (HGNNs) have shown promise in preserving network hierarchy. However, they do not model the notion of topics, thus document representations lack semantic interpretability. On the other hand, a corpus of documents usually has high variability in degrees of topic specificity. For example, some documents contain general content (e.g., sports), while others focus on specific themes (e.g., basketball and swimming). Topic models indeed model latent topics for semantic interpretability, but most assume a flat topic structure and ignore such semantic hierarchy. Given these two challenges, we propose a Hyperbolic Graph Topic Modeling Network to integrate both network hierarchy across linked documents and semantic hierarchy within texts into a unified HGNN framework. Specifically, we construct a two-layer document graph. Intra- and cross-layer encoding captures network hierarchy. We design a topic tree for text decoding to preserve semantic hierarchy and learn interpretable topics. Supervised and unsupervised experiments verify the effectiveness of our model.

A recommender system operates in a dynamic environment where new items emerge and new users join the system, resulting in ever-growing user-item interactions over time. Existing works either assume a model trained offline on a static dataset (requiring periodic re-training with ever larger datasets); or an online learning setup that favors recency over history. As privacy-aware users could hide their histories, the loss of older information means that periodic retraining may not always be feasible, while online learning may lose sight of users’ long-term preferences. In this work, we adopt a continual learning perspective to collaborative filtering, by compartmentalizing users and items over time into a notion of tasks. Of particular concern is to mitigate catastrophic forgetting that occurs when the model would reduce performance for older users and items in prior tasks even as it tries to fit the newer users and items in the current task. To alleviate this, we propose a method that leverages gradient alignment to deliver a model that is more compatible across tasks and maximizes user agreement for better user representations to improve long-term recommendations.

In the richly multimedia Web, detecting sentiment signals expressed in images would support multiple applications, e.g., measuring customer satisfaction from online reviews, analyzing trends and opinions from social media. Given an image, visual sentiment analysis aims at recognizing positive or negative sentiment, and occasionally neutral sentiment as well. A nascent yet promising direction is Transformer-based models applied to image data, whereby Vision Transformer (ViT) establishes remarkable performance on large-scale vision benchmarks. In addition to investigating the fitness of ViT for visual sentiment analysis, we further incorporate concept orientation into the self-attention mechanism, which is the core component of Transformer. The proposed model captures the relationships between image features and specific concepts. We conduct extensive experiments on Visual Sentiment Ontology (VSO) and Yelp.com online review datasets, showing that not only does the proposed model significantly improve upon the base model ViT in detecting visual sentiment but it also outperforms previous visual sentiment analysis models with narrowly-defined orientations. Additional analyses yield insightful results and better understanding of the concept-oriented self-attention mechanism.

This article introduces a novel architecture for two objectives recommendation and interpretability in a unified model. We leverage textual content as a source of interpretability in content-aware recommender systems. The goal is to characterize user preferences with a set of human-understandable attributes, each is described by a single word, enabling comprehension of user interests behind item adoptions. This is achieved via a dedicated architecture, which is interpretable by design, involving two components for recommendation and interpretation. In particular, we seek an interpreter , which accepts holistic user’s representation from a recommender to output a set of activated attributes describing user preferences. Besides encoding interpretability properties such as fidelity, conciseness and diversity, the proposed memory network-based interpreter enables the generalization of user representation by discovering relevant attributes that go beyond her adopted items’ textual content. We design experiments involving both human- and functionally-grounded evaluations of interpretability. Results on four real-world datasets show that our proposed model not only discovers highly relevant attributes for interpreting user preferences, but also enjoys comparable or better recommendation accuracy than a series of baselines.

A fundamental problem in many scenarios is to match entities across two data sources. It is frequently presumed in prior work that entities to be matched are of comparable granularity. In this work, we address one-to-many or poly-matching in the scenario where entities have varying granularity. A distinctive feature of our problem is its bidirectional nature, where the ‘one’ or the ‘many’ could come from either source arbitrarily. Moreover, to deal with diverse entity representations that give rise to noisy similarity values, we incorporate novel notions of receptivity and reclusivity into a robust matching objective. As the optimal solution to the resulting formulation is proven computationally intractable, we propose more scalable yet still performant heuristics. Experiments on multiple real-life datasets showcase the effectiveness and outperformance of our proposed algorithms over baselines.

Recommendation explanations help to improve their acceptance by end users. The form of explanation of interest here is presenting an existing review of the recommended item. The challenge is in selecting a suitable review, which is customarily addressed by assessing the relative importance of each review to the recommendation objective. Our focus is on improving review-level explanation by leveraging additional information in the form of questions and answers (QA). The proposed framework employs QA in an attention mechanism that aligns reviews to various QAs of an item and assesses their contribution jointly to the recommendation objective. The benefits are two-fold. For one, QA aids in selecting more useful reviews. For another, QA itself could accompany a well-aligned review in an expanded form of explanation. Experiments showcase the efficacies of our method as compared to baselines in identifying useful reviews and QAs, while maintaining parity in recommendation performance.

We examine the bundle configuration problem in the presence of competition. Given a competitor’s bundle configuration and pricing, we determine what to bundle together, and at what prices, to maximize the target firm’s revenue. We highlight the difficulty in pricing bundles and propose a scalable alternative and an efficient search heuristic to refine the approximate prices. Furthermore, we extend the heuristics proposed by previous work to accommodate the presence of a competitor. We analyze the effectiveness of our proposed models through experimentation on real-life ratings-based preference data.

Classical recommendation methods typically render user representation as a single vector in latent space. Oftentimes, a user's interactions with items are influenced by several hidden factors. To better uncover these hidden factors, we seek disentangled representations. Existing disentanglement methods for recommendations are mainly concerned with user-item interactions alone. To further improve not only the effectiveness of recommendations but also the interpretability of the representations, we propose to learn a second set of disentangled user representations from textual content and to align the two sets of representations with one another. The purpose of this coupling is two-fold. For one benefit, we leverage textual content to resolve sparsity of user-item interactions, leading to higher recommendation accuracy. For another benefit, by regularizing factors learned from user-item interactions with factors learned from textual content, we map uninterpretable dimensions from user representation into words. An attention-based alignment is introduced to align and enrich hidden factors representations. A series of experiments conducted on four real-world datasets show the efficacy of our methods in improving recommendation quality.

While Variational Graph Auto-Encoder (VGAE) has presented promising ability to learn representations for documents, most existing VGAE methods do not model a latent topic structure and therefore lack semantic interpretability. Exploring hidden topics within documents and discovering key words associated with each topic allow us to develop a semantic interpretation of the corpus. Moreover, documents are usually associated with authors. For example, news reports have journalists specializing in writing certain type of events, academic papers have authors with expertise in certain research topics, etc. Modeling authorship information could benefit topic modeling, since documents by the same authors tend to reveal similar semantics. This observation also holds for documents published on the same venues. However, most topic models ignore the auxiliary authorship and publication venues. Given above two challenges, we propose a Variational Graph Author Topic Model for documents to integrate both semantic interpretability and authorship and venue modeling into a unified VGAE framework. For authorship and venue modeling, we construct a hierarchical multi-layered document graph with both intra- and cross-layer topic propagation. For semantic interpretability, three word relations (contextual, syntactic, semantic) are modeled and constitute three word sub-layers in the document graph. We further propose three alternatives for variational divergence. Experiments verify the effectiveness of our model on supervised and unsupervised tasks.

In many visually-oriented applications, users can select and group images that they find interesting into coherent clusters. For instance, we encounter these in the form of hashtags on Instagram, galleries on Flickr, or boards on Pinterest. The selection and coherence of such user-curated visual clusters arise from a user's preference for a certain type of content as well as her own perception of which images are similar and thus belong to a cluster. We seek to model such curation behaviors towards supporting users in their future activities such as expanding existing clusters or discovering new clusters altogether. This paper proposes a framework, namely COLLABORATIVE CURATING that jointly models the interrelated modalities of preference expression and similarity perception. Extensive experiments on real-world datasets from a visual curating platform show that the proposed framework significantly outperforms baselines focusing on either clustering behaviors or preferences alone.

In e-commerce websites, multiple related product recommendations are usually organized into “widgets”, each given a name, as a recommendation caption, to describe the products within. These recommendation captions are usually manually crafted and generic in nature, making it difficult to attach meaningful and informative names at scale. As a result, the captions are inadequate in helping customers to better understand the connection between the multiple recommendations and make faster product discovery.

Recommender Systems ( RecSys ) provide suggestions in many decision-making processes. Given that groups of people can perform many real-world activities (e.g., a group of people attending a conference looking for a place to dine), the need for recommendations for groups has increased. A wide range of Group Recommender Systems ( GRecSys ) has been developed to aggregate individual preferences to group preferences. We analyze 175 studies related to GRecSys . Previous works evaluate their systems using different types of groups (sizes and cohesiveness), and most of such works focus on testing their systems using only one type of item, called Experience Goods (EG). As a consequence, it is hard to get consistent conclusions about the performance of GRecSys . We present the aggregation strategies and aggregation functions that GRecSys commonly use to aggregate group members’ preferences. This study experimentally compares the performance (i.e., accuracy, ranking quality, and usefulness) using four metrics (Hit Ratio, Normalize Discounted Cumulative Gain, Diversity, and Coverage) of eight representative RecSys for group recommendations on ephemeral groups. Moreover, we use two different aggregation strategies, 10 different aggregation functions, and two different types of items on two types of datasets (EG and Search Goods (SG)) containing real-life datasets. The results show that the evaluation of GRecSys needs to use both EG and SG types of data, because the different characteristics of datasets lead to different performance. GRecSys using Singular Value Decomposition or Neural Collaborative Filtering methods work better than others. It is observed that the Average aggregation function is the one that produces better results.

As a well-established probabilistic method, topic models seek to uncover latent semantics from plain text. In addition to having textual content, we observe that documents are usually compared in listwise rankings based on their content. For instance, world-wide countries are compared in an international ranking in terms of electricity production based on their national reports. Such document comparisons constitute additional information that reveal documents' relative similarities. Incorporating them into topic modeling could yield comparative topics that help to differentiate and rank documents. Furthermore, based on different comparison criteria, the observed document comparisons usually cover multiple aspects, each expressing a distinct ranked list. For example, a country may be ranked higher in terms of electricity production, but fall behind others in terms of life expectancy or government budget. Each comparison criterion, or aspect, observes a distinct ranking. Considering such multiple aspects of comparisons based on different ranking criteria allows us to derive one set of topics that inform heterogeneous document similarities. We propose a generative topic model aimed at learning topics that are well aligned to multi-aspect listwise comparisons. Experiments on public datasets demonstrate the advantage of the proposed method in jointly modeling topics and ranked lists against baselines comprehensively.

Entity matching across two data sources is a prevalent need in many domains, including e-commerce. Of interest is the scenario where entities have varying granularity, e.g., a coarse product category may match multiple finer categories. Previous work in one-to-many matching generally presumes the ‘one’ necessarily comes from a designated source and the ‘many’ from the other source. In contrast, we propose a novel formulation that allows concurrent one-to-many bidirectional matching in any direction. Beyond flexibility, we also seek matching that is more robust to noisy similarity values arising from diverse entity descriptions, by introducing receptivity and reclusivity notions. In addition to an optimal formulation, we also propose an efficient and performant heuristic. Experiments on multiple real-life datasets from e-commerce sources showcase the effectiveness and outperformance of our proposed algorithms over baselines.

Data sparsity is a long-standing challenge in recommender systems. Among existing approaches to alleviate this problem, cross-domain recommendation consists in leveraging knowledge from a source domain or category (e.g., Movies) to improve item recommendation in a target domain (e.g., Books). In this work, we advocate a probabilistic approach to cross-domain recommendation and rely on variational autoencoders (VAEs) as our latent variable models. More precisely, we assume that we have access to a VAE trained on the source domain that we seek to leverage to improve preference modeling in the target domain. To this end, we propose a model which learns to fit the target observations and align its hidden space with the source latent space jointly. Since we model the latent spaces by the variational posteriors, we operate at this level, and in particular, we investigate two approaches, namely rigid and soft alignments. In the former scenario, the variational model in the target domain is set equal to the source variational model. That is, we only learn a generative model in the target domain. In the soft-alignment scenario, the target VAE has its variational model, but which is encouraged to look like its source counterpart. We analyze the proposed objectives theoretically and conduct extensive experiments to illustrate the benefit of our contribution. Empirical results on six real-world datasets show that the proposed models outperform several comparable cross-domain recommendation models.

Recommender systems typically learn from user-item preference data such as ratings and clicks. This information is sparse in nature, i.e., observed user-item preferences often represent less than 5% of possible interactions. One promising direction to alleviate data sparsity is to leverage auxiliary information that may encode additional clues on how users consume items. Examples of such data (referred to as modalities) are social networks, item’s descriptive text, product images. The objective of this tutorial is to offer a comprehensive review of recent advances to represent, transform and incorporate the different modalities into recommendation models. Moreover, through practical hands-on sessions, we consider cross model/modality comparisons to investigate the importance of different methods and modalities. The hands-on exercises are conducted with Cornac (https://cornac.preferred.ai ), a comparative framework for multimodal recommender systems. The materials are made available on https://preferred.ai/recsys21-tutorial/.

To aid users in choice-making, explainable recommendation models seek to provide not only accurate recommendations but also accompanying explanations that help to make sense of those recommendations. Most of the previous approaches rely on evaluative explanations, assessing the quality of an individual item along some aspects of interest to the user. In this work, we are interested in comparative explanations, the less studied problem of assessing a recommended item in comparison to another reference item.

Preference data is a form of dyadic data, with measurements associated with pairs of elements arising from two discrete sets of objects. These are users and items, as well as their interactions, e.g., ratings. We are interested in learning representations for both sets of objects, i.e., users and items, to predict unknown pairwise interactions. Motivated by the recent successes of deep latent variable models, we propose Bilateral Variational Autoencoder (BiVAE), which arises from a combination of a generative model of dyadic data with two inference models, user- and item-based, parameterized by neural networks. Interestingly, our model can take the form of a Bayesian variational autoencoder either on the user or item side. As opposed to the vanilla VAE model, BiVAE is "bilateral'', in that users and items are treated similarly, making it more apt for two-way or dyadic data. While theoretically sound, we formally show that, similarly to VAE, our model might suffer from an over-regularized latent space. This issue, known as posterior collapse in the VAE literature, may appear due to assuming an over-simplified prior (isotropic Gaussian) over the latent space. Hence, we further propose a mitigation of this issue by introducing constrained adaptive prior (CAP) for learning user- and item-dependent prior distributions. Empirical results on several real-world datasets show that the proposed model outperforms conventional VAE and other comparative collaborative filtering models in terms of item recommendation. Moreover, the proposed CAP further boosts the performance of BiVAE. An implementation of BiVAE is available on Cornac recommender library.

Key to recommender systems is learning user preferences, which are expressed through various modalities. In online reviews, for instance, this manifests in numerical rating, textual content, as well as visual images. In this work, we hypothesize that modelling these modalities jointly would result in a more holistic representation of a review towards more accurate recommendations. Therefore, we propose Multimodal Review Generation (MRG), a neural approach that simultaneously models a rating prediction component and a review text generation component. We hypothesize that the shared user and item representations would augment the rating prediction with richer information from review text, while sensitizing the generated review text to sentiment features based on user and item of interest. Moreover, when review photos are available, visual features could inform the review text generation further. Comprehensive experiments on real-life datasets from several major US cities show that the proposed model outperforms comparable multimodal baselines, while an ablation analysis establishes the relative contributions of the respective components of the joint model.

Determining the similarity between two objects is pertinent to many applications. When the basis for similarity is a set of object-to-object relationships, it is natural to rely on graph-theoretic measures. One seminal technique for measuring the structural-context similarity between a pair of graph vertices is SimRank, whose underlying intuition is that two objects are similar if they are connected by similar objects. However, by design, SimRank as well as its variants capture only a single view or perspective of similarity. Meanwhile, in many real-world scenarios, there emerge multiple perspectives of similarity, i.e., two objects may be similar from one perspective, but dissimilar from another. For instance, human subjects may generate varied, yet valid, clusterings of objects. In this work, we propose a graph-theoretic similarity measure that is natively multiperspective. In our approach, the observed object-to-object relationships due to various perspectives are integrated into a unified graph-based representation, stylised as a hypergraph to retain the distinct perspectives. We then introduce a novel model for learning and reflecting diverse similarity perceptions given the hypergraph, yielding the similarity score between any pair of objects from any perspective. In addition to proposing an algorithm for computing the similarity scores, we also provide theoretical guarantees on the convergence of the algorithm. Experiments on public datasets show that the proposed model deals better with multiperspectivity than the baselines.

Top-k recommendation seeks to deliver a personalized recommendation list of k items to a user. The dual objectives are (1) accuracy in identifying the items a user is likely to prefer, and (2) efficiency in constructing the recommendation list in real time. One direction towards retrieval efficiency is to formulate retrieval as approximate k nearest neighbor (kNN) search aided by indexing schemes, such as locality-sensitive hashing, spatial trees, and inverted index. These schemes, applied on the output representations of recommendation algorithms, speed up the retrieval process by automatically discarding a large number of potentially irrelevant items when given a user query vector. However, many previous recommendation algorithms produce representations that may not necessarily align well with the structural properties of these indexing schemes, eventually resulting in a significant loss of accuracy post-indexing. In this paper, we introduce Indexable Bayesian Personalized Ranking (IBPR) that learns from ordinal preference to produce representation that is inherently compatible with the aforesaid indices. Experiments on publicly available datasets show superior performance of the proposed model compared to state-of-the-art methods on top-k recommendation retrieval task, achieving significant speedup while maintaining high accuracy.

Exploratory analysis of a text corpus is an important task that can be aided by informative visualization. One spatially-oriented form of document visualization is a scatterplot, whereby every document is associated with a coordinate, and relationships among documents can be perceived through their spatial distances. Semantic visualization further infuses the visualization space with latent semantics, by incorporating a topic model that has a representation in the visualization space, allowing users to also perceive relationships between documents and topics spatially. We illustrate how a semantic visualization system called SemVis could be used to navigate a text corpus interactively and topically via browsing and searching.

Personalized recommendation of items frequently faces scenarios where we have sparse observations on users' adoption of items. In the literature, there are two promising directions. One is to connect sparse items through similarity in content. The other is to connect sparse users through similarity in social relations. We seek to integrate both types of information, in addition to the adoption information, within a single integrated model. Our proposed method models item content via a topic model, and user communities via an autoencoder model, while bridging a user's community-based preference to her topic-based preference. Experiments on public real-life data showcase the utility of the model, particularly when there is significant compatibility between communities and topics.

Online reviews are important decision aids to consumers. Other than helping users to evaluate individual products, reviews also support comparison shopping by comparing two (or more) products based on a specific aspect. However, making a comparison across two different reviews, written by different authors, is not always equitable due to the different standards and preferences of authors. Therefore, we focus on comparative sentences, whereby two products are compared directly by a review author within a sentence. We study the problem of comparative relation mining. Given a set of comparative sentences, each relating a pair of entities, our objective is three-fold: to interpret the comparative direction in each sentence, to identify the aspect of each sentence, and to determine the relative merits of each entity with respect to that aspect. This requires mining comparative relations at two levels of resolution: at the sentence level, and at the entity level. Our insight is that there is a significant synergy between the two levels. We propose a generative model for comparative text, which jointly models comparative directions at the sentence level, and ranking at the entity level. This model is tested comprehensively on Amazon reviews dataset with good empirical outperformance over pipelined baselines.

Learning to rank is an important problem in many scenarios, such as information retrieval, natural language processing, recommender systems, etc. The objective is to learn a function that ranks a number of instances based on their features. In the vast majority of the learning to rank literature, there is an implicit assumption that the population of ranking instances are homogeneous, and thus can be modeled by a single central ranking function. In this work, we are concerned with learning to rank for a heterogeneous population, which may consist of a number of sub-populations, each of which may rank objects differently. Because these sub-populations are not known in advance, and are effectively latent, the problem turns into simultaneously learning both a set of ranking functions, as well as the latent assignment of instances to functions. To address this problem in a joint manner, we develop a probabilistic graphical model called Plackett-Luce Regression Mixture or PLRM model, and describe its inference via Expectation-Maximization algorithm. Comprehensive experiments on publicly-available real-life datasets showcase the effectiveness of PLRM, as opposed to a pipelined approach of clustering followed by learning to rank, as well as approaches that assume a single ranking function for a heterogeneous population.

Users express their personal preferences through ratings, adoptions, and other consumption behaviors. We seek to learn latent representations for user preferences from such behavioral data. One representation learning model that has been shown to be effective for large preference datasets is Restricted Boltzmann Machine (RBM). While homophily, or the tendency of friends to share their preferences at some level, is an established notion in sociology, thus far it has not yet been clearly demonstrated on RBM-based preference models. The question lies in how to appropriately incorporate social network into the architecture of RBM-based models for learning representations of preferences. In this paper, we propose two potential architectures: one that models social network among users as additional observations, and another that incorporates social network into the sharing of hidden units among related users. We study the efficacies of these proposed architectures on publicly available, real-life preference datasets with social networks, yielding useful insights.

Embedding deals with reducing the high-dimensional representation of data into a low-dimensional representation. Previous work mostly focuses on preserving similarities among objects. Here, not only do we explicitly recognize multiple types of objects, but we also focus on the ordinal relationships across types. Collaborative Ordinal Embedding or COE is based on generative modelling of ordinal triples. Experiments show that COE outperforms the baselines on objective metrics, revealing its capacity for information preservation for ordinal data.

The long-tail theory for consumer demand implies the need for more accurate personalization technologies to target items to the users who most desire them. A key tenet of personalization is the capacity to model user preferences. Most of the previous work on recommendation and personalization has focused primarily on individual preferences. While some focus on shared preferences between pairs of users, they assume that the same similarity value applies to all items. Here we investigate the notion of “context,” hypothesizing that while two users may agree on their preferences on some items, they may also disagree on other items. To model this, we design probabilistic models for the generation of rating differences between pairs of users across different items. Since this model also involves the estimation of rating differences on unseen items for the purpose of prediction, we further conduct a systematic analysis of matrix factorization and tensor factorization methods in this estimation, and propose a factorization model with a novel objective function of minimizing error in rating differences. Experiments on several real-life rating datasets show that our proposed model consistently yields context-specific similarity values that perform better on a prediction task than models relying on shared preferences.

Multipartite entity resolution seeks to match entity mentions across several collections. An entity mention is presumed unique within a collection, and thus could match at most one entity mention in each of the other collections. In addition to domain-specific features considered in entity resolution, there are a number of domain-invariant structural contraints that apply in this scenario, including one-to-one assignment as well as cross-collection transitivity. We propose a principled solution to the multipartite entity resolution problem, building on the foundation of Markov Logic Network (MLN) that combines probabilistic graphical model and first-order logic. We describe how the domain-invariant structural constraints could be expressed appropriately in terms of Markov logic, flexibly allowing joint modeling with domain-specific features. Experiments on two real-life datasets, each spanning four collections, show the utility of this approach and validate the contributions of various MLN components.

Micro-reviews is a new type of user-generated content arising from the prevalence of mobile devices and social media in the past few years. Micro-reviews are bite-size reviews (usually under 200 characters), commonly posted on social media or check-in services, using a mobile device. They capture the immediate reaction of users, and they are rich in information, concise, and to the point. However, the abundance of micro-reviews, and their telegraphic nature make it increasingly difficult to go through them and extract the useful information, especially on a mobile device. In this paper, we address the problem of summarizing the micro-reviews of an entity, such that the summary is representative, compact, and readable. We formulate the summarization problem as that of synthesizing a new ``review'' using snippets of full-text reviews. To produce a summary that naturally balances compactness and representativeness, we work within the Minimum Description Length framework. We show that finding the optimal summary is NP-hard, and we consider approximation and heuristic algorithms. We perform a thorough evaluation of our methodology on real-life data collected from Foursquare and Yelp. We demonstrate that our summaries outperform individual reviews, as well as existing summarization approaches.

With greater prevalence of social media, there is an increasing amount of user-generated data revealing consumer preferences for various products and services. Businesses seek to harness this wealth of data to improve their marketing strategies. Bundling, or selling two or more items for one price is a highly-practiced marketing strategy. In this paper, we address the bundle configuration problem from the data-driven perspective. Given a set of items in a seller's inventory, we seek to determine which items should belong to which bundle so as to maximize the total revenue, by mining consumer preferences data. We show that this problem is NP-hard when bundles are allowed to contain more than two items. Therefore, we describe an optimal solution for bundle sizes up to two items, and propose two heuristic solutions for bundles of any larger size. We investigate the effectiveness and the efficiency of the proposed algorithms through experimentations on real-life rating-based preferences data.

Given the proliferation of review content, and the fact that reviews are highly diverse and often unnecessarily verbose, users frequently face the problem of selecting the appropriate reviews to consume. Micro-reviews are emerging as a new type of online review content in the social media. Micro-reviews are posted by users of check-in services such as Foursquare. They are concise (up to 200 characters long) and highly focused, in contrast to the comprehensive and verbose reviews. In this paper, we propose a novel mining problem, which brings together these two disparate sources of review content. Specifically, we use coverage of micro-reviews as an objective for selecting a set of reviews that cover efficiently the salient aspects of an entity. Our approach consists of a two-step process: matching review sentences to micro-reviews, and selecting a small set of reviews that cover as many micro-reviews as possible, with few sentences. We formulate this objective as a combinatorial optimization problem, and show how to derive an optimal solution using Integer Linear Programming. We also propose an efficient heuristic algorithm that approximates the optimal solution. Finally, we perform a detailed evaluation of all the steps of our methodology using data collected from Foursquare and Yelp.

We propose CCRank, the first parallel framework for learning to rank based on evolutionary algorithms (EA), aiming to significantly improve learning efficiency while maintaining accuracy. CCRank is based on cooperative coevolution (CC), a divide-and-conquer framework that has demonstrated high promise in function optimization for problems with large search space and complex structures. Moreover, CC naturally allows parallelization of sub-solutions to the decomposed sub-problems, which can substantially boost learning efficiency. With CCRank, we investigate parallel CC in the context of learning to rank. We implement CCRank with three EA-based learning to rank algorithms for demonstration. Extensive experiments on benchmark datasets in comparison with the state-of-the-art algorithms show the performance gains of CCRank in efficiency and accuracy.

With the prevalence of the Web and social media, users increasingly express their preferences online. In learning these preferences, recommender systems need to balance the trade-off between exploitation, by providing users with more of the "same", and exploration, by providing users with something "new" so as to expand the systems' knowledge. Multi-armed bandit (MAB) is a framework to balance this trade-off. Most of the previous work in MAB either models a single bandit for the whole population, or one bandit for each user. We propose an algorithm to divide the population of users into multiple clusters, and to customize the bandits to each cluster. This clustering is dynamic, i.e., users can switch from one cluster to another, as their preferences change. We evaluate the proposed algorithm on two real-life datasets.

Users frequently rely on online reviews for decision making. In addition to allowing users to evaluate the quality of individual products, reviews also support comparison shopping. One key user activity is to compare two (or more) products based on a specific aspect. However, making a comparison across two different reviews, written by different authors, is not always equitable due to the different standards and preferences of individual authors. Therefore, we focus instead on comparative sentences, whereby two products are compared directly by a review author within a single sentence.

A document network refers to a data type that can be represented as a graph of vertices, where each vertex is associated with a text document. Examples of such a data type include hyperlinked Web pages, academic publications with citations, and user profiles in social networks. Such data have very high-dimensional representations, in terms of text as well as network connectivity. In this paper, we study the problem of embedding, or finding a low-dimensional representation of a document network that "preserves" the data as much as possible. These embedded representations are useful for various applications driven by dimensionality reduction, such as visualization or feature selection. While previous works in embedding have mostly focused on either the textual aspect or the network aspect, we advocate a holistic approach by finding a unified low-rank representation for both aspects. Moreover, to lend semantic interpretability to the low-rank representation, we further propose to integrate topic modeling and embedding within a joint model. The gist is to join the various representations of a document (words, links, topics, and coordinates) within a generative model, and to estimate the hidden representations through MAP estimation. We validate our model on real-life document networks, showing that it outperforms comparable baselines comprehensively on objective evaluation metrics.

Visualization of high-dimensional data such as text documents is widely applicable. The traditional means is to find an appropriate embedding of the high-dimensional representation in a low-dimensional visualizable space. As topic modeling is a useful form of dimensionality reduction that preserves the semantics in documents, recent approaches aim for a visualization that is consistent with both the original word space, as well as the semantic topic space. In this paper, we address the semantic visualization problem. Given a corpus of documents, the objective is to simultaneously learn the topic distributions as well as the visualization coordinates of documents. We propose to develop a semantic visualization model that approximates L2-normalized data directly. The key is to associate each document with three representations: a coordinate in the visualization space, a multinomial distribution in the topic space, and a directional vector in a high-dimensional unit hypersphere in the word space. We join these representations in a unified generative model, and describe its parameter estimation through variational inference. Comprehensive experiments on real-life text datasets show that the proposed method outperforms the existing baselines on objective evaluation metrics for visualization quality and topic interpretability.

Personalization, or customizing the experience of each individual user, is seen as a useful way to navigate the huge variety of choices on the Web today. A key tenet of personalization is the capacity to model user preferences. The paradigm has shifted from that of individual preferences, whereby we look at a user's past activities alone, to that of shared preferences, whereby we model the similarities in preferences between pairs of users (e.g., friends, people with similar interests). However, shared preferences are still too granular, because it assumes that a pair of users would share preferences across all items. We therefore postulate the need to pay attention to "context", which refers to the specific item on which the preferences between two users are to be estimated. In this paper, we propose a generative model for contextual agreement in preferences. For every triplet consisting of two users and an item, the model estimates both the prior probability of agreement between the two users, as well as the posterior probability of agreement with respect to the item at hand. The model parameters are estimated from ratings data. To extend the model to unseen ratings, we further propose several matrix factorization techniques focused on predicting agreement, rather than ratings. Experiments on real-life data show that our model yields context-specific similarity values that perform better on a prediction task than models relying on shared preferences.

Online reviews are an invaluable resource for web users trying to make decisions regarding products or services. However, the abundance of review content, as well as the unstructured, lengthy, and verbose nature of reviews make it hard for users to locate the appropriate reviews, and distill the useful information. With the recent growth of social networking and micro-blogging services, we observe the emergence of a new type of online review content, consisting of bite-sized, 140 character-long reviews often posted reactively on the spot via mobile devices. These micro-reviews are short, concise, and focused, nicely complementing the lengthy, elaborate, and verbose nature of full-text reviews.

Graph-based proximity has many applications with different ranking needs. However, most previous works only stress the sense of importance by finding “popular” results for a query. Often times important results are overly general without being well-tailored to the query, lacking a sense of specificity - which only emerges recently. Even then, the two senses are treated independently, and only combined empirically. In this paper, we generalize the well-studied importance-based random walk into a round trip and develop RoundTripRank, seamlessly integrating specificity and importance in one coherent process. We also recognize the need for a flexible trade-off between the two senses, and further develop RoundTripRank+ based on a scheme of hybrid random surfers. For efficient computation, we start with a basic model that decomposes RoundTripRank into smaller units. For each unit, we apply a novel two-stage bounds updating framework, enabling an online top-K algorithm 2SBound. Finally, our experiments show that RoundTripRank and RoundTripRank+ are robust over various ranking tasks, and 2SBound enables scalable online processing.

User preferences are commonly learned from historical data whereby users express preferences for items, e.g., through consumption of products or services. Most work assumes that a user is not constrained in their selection of items. This assumption does not take into account the availability constraint, whereby users could only access some items, but not others. For example, in subscription-based systems, we can observe only those historical preferences on subscribed (available) items. However, the objective is to predict preferences on unsubscribed (unavailable) items, which do not appear in the historical observations due to their (lack of) availability. To model preferences in a probabilistic manner and address the issue of availability constraint, we develop a graphical model, called Latent Transition Model (LTM) to discover users' latent interests. LTM is novel in incorporating transitions in interests when certain items are not available to the user. Experiments on a real-life implicit feedback dataset demonstrate that LTM is effective in discovering customers' latent interests, and it achieves significant improvements in prediction accuracy over baselines that do not model transitions.

Users face many choices on the web when it comes to choosing which product to buy, which video to watch, and so on. In making adoption decisions, users rely not only on their own preferences, but also on friends. We call the latter social correlation, which may be caused by the homophily and social influence effects. In this paper, we focus on modeling social correlation on users item adoptions. Given a user-user social graph and an item-user adoption graph, our research seeks to answer the following questions: Whether the items adopted by a user correlate with items adopted by her friends, and how to model item adoptions using social correlation. We propose a social correlation framework that considers a social correlation matrix representing the degrees of correlation from every user to the users friends, in addition to a set of latent factors representing topics of interests of individual users. Based on the framework, we develop two generative models, namely sequential and unified, and the corresponding parameter estimation approaches. From each model, we devise the social correlation only and hybrid methods for predicting missing adoption links. Experiments on LiveJournal and Epinions data sets show that our proposed models outperform the approach based on latent factors only (LDA).

User-to-user interactions have become ubiquitous in Web 2.0. Users exchange emails, post on newsgroups, tag web pages, co-author papers, etc. Through these interactions, users co-produce or co-adopt content items (e.g., words in emails, tags in social bookmarking sites). We model such dynamic interactions as a user interaction network, which relates users, interactions, and content items over time. After some interactions, a user may produce content that is more similar to those produced by other users previously. We term this effect social dependency, and we seek to mine from such networks the degree to which a user may be socially dependent on another user over time. We propose a Decay Topic Model to model the evolution of a user's preferences for content items at the topic level, as well as a Social Dependency Metric that quantifies the extent of social dependency based on interactions and content changes. Our experiments on two user interaction networks induced from real-life datasets show the effectiveness of our approach.

Nowadays, there are many queries issued to search engines targeting at finding values from structured data (e.g., movie showtime of a specific location). In such scenarios, there is often a mismatch between the values of structured data (how content creators describe entities) and the web queries (how different users try to retrieve them). Therefore, recognizing the alternative ways people use to reference an entity, is crucial for structured web search. In this paper, we study the problem of automatic generation of entity synonyms over structured data toward closing the gap between users and structured data. We propose an offline, data-driven approach that mines query logs for instances where content creators and web users apply a variety of strings to refer to the same webpages. This way, given a set of strings that reference entities, we generate an expanded set of equivalent strings (entity synonyms) for each entity. Our framework consists of three modules: candidate generation, candidate selection, and noise cleaning. We further study the cause of the problem through the identification of different entity synonym classes. The proposed method is verified with experiments on real-life data sets showing that we can significantly increase the coverage of structured web queries with good precision.

Users are increasingly pursuing complex task-oriented goals on the web, such as making travel arrangements, managing finances, or planning purchases. To this end, they usually break down the tasks into a few codependent steps and issue multiple queries around these steps repeatedly over long periods of time. To better support users in their long-term information quests on the web, search engines keep track of their queries and clicks while searching online. In this paper, we study the problem of organizing a user's historical queries into groups in a dynamic and automated fashion. Automatically identifying query groups is helpful for a number of different search engine components and applications, such as query suggestions, result ranking, query alterations, sessionization, and collaborative search. In our approach, we go beyond approaches that rely on textual similarity or time thresholds, and we propose a more robust approach that leverages search query logs. We experimentally study the performance of different techniques, and showcase their potential, especially when combined together.

The emerging trend of social information processing has resulted in Web users’ increased reliance on user-generated content contributed by others for information searching and decision making. Rating scores, a form of user-generated content contributed by reviewers in online rating systems, allow users to leverage others’ opinions in the evaluation of objects. In this article, we focus on the problem of summarizing the rating scores given to an object into an overall score that reflects the object’s quality. We observe that the existing approaches for summarizing scores largely ignores the effect of reviewers exercising different standards in assigning scores. Instead of treating all reviewers as equals, our approach models the leniency of reviewers, which refers to the tendency of a reviewer to assign higher scores than other coreviewers. Our approach is underlined by two insights: (1) The leniency of a reviewer depends not only on how the reviewer rates objects, but also on how other reviewers rate those objects and (2) The leniency of a reviewer and the quality of rated objects are mutually dependent. We develop the leniency-aware quality , or LQ model, which solves leniency and quality simultaneously. We introduce both an exact and a ranked solution to the model. Experiments on real-life and synthetic datasets show that LQ is more effective than comparable approaches. LQ is also shown to perform consistently better under different parameter settings.

Users face a dazzling array of choices on the Web when it comes to choosing which product to buy, which video to watch, etc. The trend of social information processing means users increasingly rely not only on their own preferences, but also on friends when making various adoption decisions. In this paper, we investigate the effects of social correlation on users' adoption of items. Given a user-user social graph and an item-user adoption graph, we seek to answer the following questions: 1) whether the items adopted by a user correlate to items adopted by her friends, and 2) how to incorporate social correlation in order to improve prediction of unobserved item adoptions. We propose the Social Correlation model based on Latent Dirichlet Allocation (LDA) that decomposes the adoption graph into a set of latent factors reflecting user preferences, and a social correlation matrix reflecting the degree of correlation from one user to another. This matrix is learned (rather than pre-assigned), has probabilistic interpretation, and preserves the underlying social network structure. We further devise a Hybrid model that combines a user's own latent factors with her friends' for adoption prediction. Experiments on Epinions and LiveJournal data sets show that our proposed models outperform the approach based on latent factors only (LDA).

Learning to rank represents a category of effective ranking methods for information retrieval. While the primary concern of existing research has been accuracy, learning efficiency is becoming an important issue due to the unprecedented availability of large-scale training data and the need for continuous update of ranking functions. In this paper, we investigate parallel learning to rank, targeting simultaneous improvement in accuracy and efficiency.

This paper aims to detect users generating spam reviews or review spammers. We identify several characteristic behaviors of review spammers and model these behaviors so as to detect the spammers. In particular, we seek to model the following behaviors. First, spammers may target specific products or product groups in order to maximize their impact. Second, they tend to deviate from the other reviewers in their ratings of products. We propose scoring methods to measure the degree of spam for each reviewer and apply them on an Amazon review dataset. We then select a subset of highly suspicious reviewers for further scrutiny by our user evaluators with the help of a web based spammer evaluation software specially developed for user evaluation experiments. Our results show that our proposed ranking and supervised methods are effective in discovering spammers and outperform other baseline method based on helpfulness votes alone. We finally show that the detected spammers have more significant impact on ratings compared with the unhelpful reviewers.

Recognizing the alternative ways people use to reference an entity, is important for many Web applications that query structured data. In such applications, there is often a mismatch between how content creators describe entities and how different users try to retrieve them. In this paper, we consider the problem of determining whether a candidate query approximately matches with an entity. We propose an off-line, data-driven, bottom-up approach that mines query logs for instances where Web content creators and Web users apply a variety of strings to refer to the same Web pages. This way, given a set of strings that reference entities, we generate an expanded set of equivalent strings for each entity. The proposed method is verified with experiments on real-life data sets showing that we can dramatically increase the queries that can be matched.

We present the design of Symphony, a platform that enables non-developers to build and deploy a new class of search-driven applications that combine their data and domain expertise with content from search engines and other web services. The Symphony prototype has been built on top of Microsoft's Bing infrastructure. While Symphony naturally makes use of the customization capabilities exposed by Bing, its distinguishing feature is the capability it provides to the application creator to combine their proprietary data and domain expertise with content obtained from Bing. They can also integrate specialized data obtained from web services to enhance the richness of their applications. Finally, Symphony is targeted at non-developers and provides cloud services for the creation and deployment of applications.

Spatio-temporal data concerning the movement of individuals over space and time contains latent information on the associations among these individuals. Sources of spatio-temporal data include usage logs of mobile and Internet technologies. This article defines a spatio-temporal event by the co-occurrences among individuals that indicate potential associations among them. Each spatio-temporal event is assigned a weight based on the precision and uniqueness of the event. By aggregating the weights of events relating two individuals, we can determine the strength of association between them. We conduct extensive experimentation to investigate both the efficacy of the proposed model as well as the computational complexity of the proposed algorithms. Experimental results on three real-life spatio-temporal datasets cross-validate each other, lending greater confidence on the reliability of our proposed model.

Wikipedia 1 is a very large and successful Web 2.0 example. As the number of Wikipedia articles and contributors grows at a very fast pace, there are also increasing disputes occurring among the contributors. Disputes often happen in articles with controversial content. They also occur frequently among contributors who are "aggressive" or controversial in their personalities. In this paper, we aim to identify controversial articles in Wikipedia. We propose three models, namely the Basic model and two Controversy Rank (CR) models. These models draw clues from collaboration and edit history instead of interpreting the actual articles or edited content. While the Basic model only considers the amount of disputes within an article, the two Controversy Rank models extend the former by considering the relationships between articles and contributors. We also derived enhanced versions of these models by considering the age of articles. Our experiments on a collection of 19,456 Wikipedia articles shows that the Controversy Rank models can more effectively determine controversial articles compared to the Basic and other baseline models

Evaluation is prevalent in real life. With the advent of Web 2.0, online evaluation has become an important feature in many applications that involve information (e.g., video, photo, and audio) sharing and social networking (e.g., blogging). In these evaluation settings, a set of reviewers assign scores to a set of objects. As part of the evaluation analysis, we want to obtain fair reviews for all the given objects. However, the reality is that reviewers may deviate in their scores assigned to the same object, due to the potential "bias" of reviewers or "controversy" of objects. The statistical approach of averaging deviations to determine bias and controversy assumes that all reviewers and objects should be given equal weight. In this paper, we look beyond this assumption and propose an approach based on the following observations: 1) evaluation is "subjective," as reviewers and objects have varying bias and controversy, respectively, and 2) bias and controversy are mutually dependent. These observations underlie our proposed reinforcement-based model to determine bias and controversy simultaneously. Our approach also quantifies "evidence," which reveals the degree of confidence with which bias and controversy have been derived. This model is shown to be effective by experiments on real-life and synthetic data sets.

Wikipedia has grown to be the world largest and busiest free encyclopedia, in which articles are collaboratively written and maintained by volunteers online. Despite its success as a means of knowledge sharing and collaboration, the public has never stopped criticizing the quality of Wikipedia articles edited by non-experts and inexperienced contributors. In this paper, we investigate the problem of assessing the quality of articles in collaborative authoring of Wikipedia. We propose three article quality measurement models that make use of the interaction data between articles and their contributors derived from the article edit history. Our B asic model is designed based on the mutual dependency between article quality and their author authority. The P eer R eview model introduces the review behavior into measuring article quality. Finally, our P rob R eview models extend P eer R eview with partial reviewership of contributors as they edit various portions of the articles. We conduct experiments on a set of well-labeled Wikipedia articles to evaluate the effectiveness of our quality measurement models in resembling human judgement.

A frequently encountered problem in decision making is the following review problem: review a large number of objects and select a small number of the best ones. An example is selecting conference papers from a large number of submissions. This problem involves two sub-problems: assigning reviewers to each object, and summarizing reviewers' scores into an overall score that supposedly reflects the quality of an object. In this paper, we address the score summarization sub-problem for the scenario where a small number of reviewers evaluate each object. Simply averaging the scores may not work as even a single reviewer could influence the average significantly. We recognize that reviewers are not necessarily on an equal ground and propose the notion of “leniency” to model this difference of reviewers. Two insights underpin our approach: (1) the “leniency” of a reviewer depends on how s/he evaluates objects as well as on how other reviewers evaluate the same set of objects, (2) the “leniency” of a reviewer and the “quality” of objects evaluated exhibit a mutual dependency relationship. These insights motivate us to develop a model that solves both “leniency” and “quality” simultaneously. We study the effectiveness of this model on a real-life dataset.

In this paper, we investigate how deviation in evaluation activities may reveal bias on the part of reviewers and controversy on the part of evaluated objects. We focus on a 'data-centric approach' where the evaluation data is assumed to represent the 'ground truth'. The standard statistical approaches take evaluation and deviation at face value. We argue that attention should be paid to the subjectivity of evaluation, judging the evaluation score not just on 'what is being said' (deviation), but also on 'who says it' (reviewer) as well as on 'whom it is said about' (object). Furthermore, we observe that bias and controversy are mutually dependent, as there is more bias if there is higher deviation on a less controversial object. To address this mutual dependency, we propose a reinforcement model to identify bias and controversy. We test our model on real-life data to verify its applicability.

Using open source Web editing software (e.g., wiki), online community users can now easily edit, review and publish articles collaboratively. While much useful knowledge can be derived from these articles, content users and critics are often concerned about their qualities. In this paper, we develop two models, namely basic model and peer review model, for measuring the qualities of these articles and the authorities of their contributors. We represent collaboratively edited articles and their contributors in a bipartite graph. While the basic model measures an article's quality using both the authorities of contributors and the amount of contribution from each contributor, the peer review model extends the former by considering the review aspect of article content. We present results of experiments conducted on some Wikipedia pages and their contributors. Our result show that the two models can effectively determine the articles' qualities and contributors' authorities using the collaborative nature of online communities