Weixin Chen, Yuhan Zhao, Jingyuan Huang, Zihe Ye, Mingxuan Ju, Tong Zhao, Neil Shah, Li Chen, Yongfeng Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Cross-domain recommendation (CDR) offers an effective strategy for improving recommendation quality in a target domain by leveraging auxiliary signals from source domains. Nonetheless, emerging evidence shows that CDR can inadvertently heighten group-level unfairness. In this work, we conduct a comprehensive theoretical and empirical analysis to uncover why these fairness issues arise. Specifically, we identify two key challenges: (i) Cross-Domain Disparity Transfer, wherein existing group-level disparities in the source domain are systematically propagated to the target domain; and (ii) Unfairness from Cross-Domain Information Gain, where the benefits derived from cross-domain knowledge are unevenly allocated among distinct groups.

Representation learning is essential for deep-neural-network-based recommender systems to capture user preferences and item features within fixed-dimensional user and item vectors. Unlike existing representation learning methods that either treat each user preference and item feature uniformly or categorize them into discrete clusters, we argue that in the real world, user preferences and item features are naturally expressed and organized in a hierarchical manner, leading to a new direction for representation learning. In this article, we introduce a novel matryoshka representation learning method for recommendation (MRL4Rec) , by which we restructure user and item vectors into matryoshka representations with nested vector spaces to explicitly represent user preferences and item features at different hierarchical layers. We theoretically establish that training with the same triplets for each sliced vector cannot guarantee representation learning with hierarchical structures. Subsequently, we propose the layer- and hardness-adaptive negative sampling (LHANS) mechanism to construct training triplets, which further ensures the soundness of learned matryoshka representations in capturing hierarchical user preferences and item features. The experiments demonstrate that MRL4Rec can consistently and substantially outperform a number of state-of-the-art competitors on several real-life datasets. Our code is publicly available at https://github.com/Riwei-HEU/MRL .

Recommender systems have been popularly employed to address information overload problems for users, for which unfairness issues are essential to be mitigated. Current user-side fairness studies in recommendations aim to ensure the independence of users’ sensitive attributes in terms of outcome or process . Specifically, the former emphasizes the outcome metric equity across user groups based on sensitive attributes, for which regularization has been proposed as a typical approach to reducing the recommendation imbalance among them. The latter process fairness focuses on independence between sensitive attributes and recommendations in the process, for which adversarial learning has been popularly adopted to remove sensitive information from user representation for individual users. However, little work has investigated the effects of these methods on balancing both group-level outcome fairness and individual-level process fairness in a particular scenario. Moreover, existing experiments have primarily been performed on one type of user attributes, e.g., behavioral attributes or demographic attributes, but neglected other possible sensitive attributes such as psychological attributes. In this article, we have investigated both outcome and process fairness performance of regularization and adversarial learning methods over multiple types of sensitive attributes, including behavioral (e.g., activity level), demographic (e.g., gender, age), and psychological attributes (e.g., big-five personality, curiosity). The experiments on four datasets show that unfairness can exist in different forms for different types of sensitive attributes, and relative to the regularization-based method, adversarial learning owns a higher potential to achieve a balance between outcome and process user-side fairness in recommendations. Our source code is available at https://github.com/WeixinChen98/OtPrFairness-MultiAttr .

Offline evaluation is the predominant method for scientific research in recommender systems, enabling the comparison of alternative recommendation approaches using pre-collected datasets and computational metrics without involving human participants. However, recent reproducibility studies reveal that many offline evaluations in the literature lack scientific rigor or adopt research practices that cast doubt on the validity of their findings. A particularly common and ultimately catastrophic flaw is the comparison of newly proposed machine learning models against untuned or poorly tuned baseline models. Combined with limited reproducibility, such practices raise serious concerns about the true progress achieved by increasingly complex recommendation algorithms. In this editorial, we argue for stronger methodological standards and summarize essential guidance and best practices for conducting rigorous offline evaluations of recommender systems. Accordingly, ACM Transactions on Recommender Systems will place increased emphasis on methodological rigor in all future submissions, with particular priority given to work that provides comprehensive reproducibility materials and clearly documents the tuning procedures used for baseline models.

Intuitively, an ideal collaborative filtering (CF) model should learn from users' full rankings over all items to make optimal top-K recommendations. Due to the absence of such full rankings in practice, most CF models rely on pairwise loss functions to approximate full rankings, resulting in an immense performance gap. In this paper, we provide a novel analysis using the multiple ordinal classification concept to reveal the inevitable gap between a pairwise approximation and the ideal case. However, bridging the gap in practice encounters two formidable challenges: (1) none of the real-world datasets contains full ranking information; (2) there does not exist a loss function that is capable of consuming ranking information. To overcome these challenges, we propose a pseudo-ranking paradigm (PRP) that addresses the lack of ranking information by introducing pseudo-rankings supervised by an original noise injection mechanism. Additionally, we put forward a new ranking loss function designed to handle ranking information effectively. To ensure our method's robustness against potential inaccuracies in pseudo-rankings, we equip the ranking loss function with a gradient-based confidence mechanism to detect and mitigate abnormal gradients. Extensive experiments on four real-world datasets demonstrate that PRP significantly outperforms state-of-the-art methods.

Cross-domain recommendation (CDR) methods predominantly leverage overlapping users to transfer knowledge from a source domain to a target domain.However, through empirical studies, we uncover a critical bias inherent in these approaches: while overlapping users experience significant enhancements in recommendation quality, non-overlapping users benefit minimally and even face performance degradation.This unfairness may erode user trust, and, consequently, negatively impact business engagement and revenue.To address this issue, we propose a novel solution that generates virtual source-domain users for non-overlapping target-domain users.Our method utilizes a dual attention mechanism to discern similarities between overlapping and non-overlapping users, thereby synthesizing realistic virtual user embeddings.We further introduce a limiter component that ensures the generated virtual users align with real-data distributions while preserving each user's unique characteristics.Notably, our method is model-agnostic and can be seamlessly integrated into any CDR model.Comprehensive experiments conducted on three public datasets with five CDR baselines demonstrate that our method effectively mitigates the CDR nonoverlapping user bias, without loss of overall accuracy.Our code is publicly available at https://github.com/WeixinChen98/VUG.

Implicit feedback has become the primary source of training data for modern recommender systems due to its abundance and ease of collection. However, the inherent noise in implicit feedback poses significant challenges to model training. Existing denoising approaches either completely remove suspected noisy interactions (re-sampling) or uniformly adjust their importance (re-weighting). Such coarse-grained treatments fail to capture the complex nature of noise in real-world scenarios, where different aspects of an interaction may have varying noise levels.

Recently, multimodal recommendations (MMRs) have gained increasing attention for alleviating the data sparsity problem of traditional recommender systems by incorporating modality-based representations. Although MMR exhibits notable improvement in recommendation accuracy, we empirically validate that an increase in the quantity or variety of modalities leads to a higher degree of users’ sensitive information leakage due to entangled causal relationships, risking fair representation learning. On the other hand, existing fair representation learning approaches are mostly based on the assumption that sensitive information is solely leaked from users’ interaction data and do not explicitly model the causal relationships introduced by multimodal data, which limits their applicability in multimodal scenarios. To address this limitation, we propose a novel fair multimodal recommendation approach (dubbed FMMRec) through causality-inspired fairness-oriented modal disentanglement and relation-aware fairness learning. Particularly, we disentangle biased and filtered modal embeddings inspired by causal inference techniques, enabling the mining of modality-based unfair and fair user–user relations, thereby enhancing the fairness and informativeness of user representations. By addressing the causal effects of sensitive attributes on user preferences, our approach aims to achieve counterfactual fairness in MMRs. Experiments on two public datasets demonstrate the superiority of our FMMRec relative to the state-of-the-art baselines. Our source code is available at https://github.com/WeixinChen98/FMMRec .

Negative sampling plays a crucial role in implicit-feedback-based collaborative filtering, where it leverages massive unlabeled data to generate negative signals for guiding supervised learning. The current state-of-the-art approaches focus on utilizing hard negative samples that contain more information to establish a better decision boundary. To strike a balance between efficiency and effectiveness, most existing methods adopt a two-pass approach: In the first pass, a fixed number of unobserved items are sampled using a simple static distribution, while, in the second pass, a more sophisticated negative sampling strategy is employed to select the final negative items. However, selecting negative samples solely from the original items in a dataset is inherently restricted due to the limited available choices and thus may not be able to effectively contrast positive samples. In this article, we empirically validate this observation through meticulously designed experiments and identify three major limitations of existing solutions: ambiguous trap, information discrimination, and false-negative samples. Our response to such limitations is to introduce “denoised” and “augmented” negative samples that may not exist in the original dataset. This direction renders a few substantial technical challenges. First, constructing augmented negative samples may introduce excessive noise that eventually distorts the decision boundary. Second, the scarcity of supervision signals hampers the denoising process. To this end, we introduce, to the best of our knowledge, a novel generic denoising and augmented negative sampling paradigm and provide a concrete instantiation. First, we disentangle the hard and easy factors of negative items. Then, we regulate the augmentation of easy factors by carefully considering the direction and magnitude. Next, we propose a reverse attention mechanism to learn a user’s negative preference, which allows us to perform a dimension-level denoising procedure on hard factors. Finally, we design an advanced negative sampling strategy to identify the final negative samples, taking into account both the score function used in existing methods and, to the best of our knowledge, a novel metric called synthesization gain. Through extensive experiments on real-world datasets, we demonstrate that our method substantially outperforms state-of-the-art baselines. Our code is publicly available at https://github.com/Asa9aoTK/ANS-Recbole .

Negative sampling is essential for implicit collaborative filtering to provide proper negative training signals so as to achieve desirable performance. We experimentally unveil a common limitation of all existing negative sampling methods that they can only select negative samples of a fixed hardness level, leading to the false positive problem (FPP) and false negative problem (FNP). We then propose a new paradigm called adaptive hardness negative sampling (AHNS) and discuss its three key criteria. By adaptively selecting negative samples with appropriate hardnesses during the training process, AHNS can well mitigate the impacts of FPP and FNP. Next, we present a concrete instantiation of AHNS called AHNS_{p

Music-based reminiscence has the potential to positively impact the psychological well-being of older adults. However, the aging process and physiological changes, such as memory decline and limited verbal communication, may impede the ability of older adults to recall their memories and life experiences. Given the advanced capabilities of generative artificial intelligence (AI) systems, such as generated conversations and images, and their potential to facilitate the reminiscing process, this study aims to explore the design of generative AI to support music-based reminiscence in older adults. This study follows a user-centered design approach incorporating various stages, including detailed interviews with two social workers and two design workshops (involving ten older adults). Our work contributes to an in-depth understanding of older adults’ attitudes toward utilizing generative AI for supporting music-based reminiscence and identifies concrete design considerations for the future design of generative AI to enhance the reminiscence experience of older adults.

Large language models (LLM) not only have revolutionized the field of natural language processing (NLP) but also have the potential to reshape many other fields, e.g., recommender systems (RS). However, most of the related work treats an LLM as a component of the conventional recommendation pipeline (e.g., as a feature extractor), which may not be able to fully leverage the generative power of LLM. Instead of separating the recommendation process into multiple stages, such as score computation and re-ranking, this process can be simplified to one stage with LLM: directly generating recommendations from the complete pool of items. This survey reviews the progress, methods, and future directions of LLM-based generative recommendation by examining three questions: 1) What generative recommendation is, 2) Why RS should advance to generative recommendation, and 3) How to implement LLM-based generative recommendation for various RS tasks. We hope that this survey can provide the context and guidance needed to explore this interesting and emerging topic.

This workshop aims to explore the evaluation and application of Large Language Models (LLMs) in recommendation systems (RSs), highlighting innovations, challenges, and future directions, focusing on enhancing RSs through LLM techniques such as prompting, fine-tuning, and developing conversational systems. By gathering researchers and partitioners from both academia and industry, the workshop focuses on discussing state-of-the-art techniques and addressing challenges and innovative applications in various sectors. At last, the workshop encourages research on topics including LLM integration, evaluating LLM-based RSs, transparency, and conversational RS development, aiming to set a research agenda for future RS advancements.

Conversational recommender systems (CRSs) aim to capture user preferences and provide personalized recommendations through interactive natural language interaction. The recent advent of large language models (LLMs) has revolutionized human engagement in natural conversation, driven by their extensive world knowledge and remarkable natural language understanding and generation capabilities. However, introducing LLMs into CRSs presents new technical challenges. Directly prompting LLMs for recommendation generation requires understanding a large and evolving item corpus, as well as grounding the generated recommendations in the real item space. On the other hand, generating recommendations based on external recommendation engines or directly integrating their suggestions into responses may constrain the overall performance of LLMs, since these engines generally have inferior representation abilities compared to LLMs. To address these challenges, we propose an end-to-end large-scale CRS model, named as ReFICR, a novel LLM-enhanced conversational recommender that empowers a retrievable large language model to perform conversational recommendation by following retrieval and generation instructions through lightweight tuning. By decomposing the complex CRS task into multiple subtasks, we formulate these subtasks into two types of instruction formats: retrieval and generation. The hidden states of ReFICR are utilized for generating text embeddings for retrieval, and simultaneously ReFICR is fine-tuned to handle generation subtasks. We optimize the contrastive objective to enhance text embeddings for retrieval and jointly fine-tune the large language model objective for generation. Our experimental results on public datasets demonstrate that ReFICR significantly outperforms baselines in terms of recommendation accuracy and response quality. Our code is publicly available at the link: https://github.com/yt556677/ReFICR.

Collaborative filtering (CF) stands as a cornerstone in recommender systems, yet effectively leveraging the massive unlabeled data presents a significant challenge. Current research focuses on addressing the challenge of unlabeled data by extracting a subset that closely approximates negative samples. Regrettably, the remaining data are overlooked, failing to fully integrate this valuable information into the construction of user preferences. To address this gap, we introduce a novel positive-neutral-negative (PNN) learning paradigm. PNN introduces a neutral class, encompassing intricate items that are challenging to categorize directly as positive or negative samples. By training a model based on this triple-wise partial ranking, PNN offers a promising solution to learning complex user preferences. Through theoretical analysis, we connect PNN to one-way partial AUC (OPAUC) to validate its efficacy. Implementing the PNN paradigm is, however, technically challenging because: (1) it is difficult to classify unlabeled data into neutral or negative in the absence of supervised signals; (2) there does not exist any loss function that can handle set-level triple-wise ranking relationships. To address these challenges, we propose a semi-supervised learning method coupled with a user-aware attention model for knowledge acquisition and classification refinement. Additionally, a novel loss function with a two-step centroid ranking approach enables handling set-level rankings. Extensive experiments on four real-world datasets demonstrate that, when combined with PNN, a wide range of representative CF models can consistently and significantly boost their performance. Even with a simple matrix factorization, PNN can achieve comparable performance to sophisticated graph neutral networks. Our code is publicly available at https://github.com/Asa9aoTK/PNN-RecBole.

An increasing number of recommendation systems try to enhance the overall user experience by incorporating conversational interaction. However, evaluating conversational recommender systems (CRSs) from the user’s perspective remains elusive. The GUI-based system evaluation criteria may be inadequate for their conversational counterparts. This article presents our proposed unifying framework, CRS-Que , to evaluate the user experience of CRSs. This new evaluation framework is developed based on ResQue , a popular user-centric evaluation framework for recommender systems. Additionally, it includes user experience metrics of conversation (e.g., understanding, response quality, humanness) under two dimensions of ResQue (i.e., Perceived Qualities and User Beliefs). Following the psychometric modeling method, we validate our framework by evaluating two conversational recommender systems in different scenarios: music exploration and mobile phone purchase . The results of the two studies support the validity and reliability of the constructs in our framework and reveal how conversation constructs and recommendation constructs interact and influence the overall user experience of the CRS. We believe this framework could help researchers conduct standardized user-centric research for conversational recommender systems and provide practitioners with insights into designing and evaluating a CRS from users’ perspectives.

Music can affect the human brain and cognition. Melodies and lyrics that resonate with us can awaken our inner feelings and thoughts; being in touch with these feelings and expressing them allow us to understand ourselves better and increase our self-awareness. To support self-awareness elicited by music, we designed a novel conversational agent (CA) that guides users to become self-aware and express their thoughts when they listen to music. Moreover, we investigated two prominent design factors in the CA, proactive guidance and social information. We then conducted a 2x2 between-subjects experiment (N = 90) to investigate how the two design factors affect self-awareness, user acceptance, and mental well-being. The results of a five-day user study reveal that high proactive guidance and social information increased self-awareness, but high proactive guidance tended to influence perceived autonomy and usefulness negatively. Further, users’ subjective feedback revealed the CA’s potential to support mental well-being.

Large language models (LLM) have manifested unparalleled modeling capability on various tasks, e.g., multi-step reasoning, but the input to these models is mostly limited to plain text, which could be very long and contain noisy information. Long text could take long time to process, and thus may not be efficient enough for recommender systems that require immediate response. In LLM-based recommendation models, user and item IDs are usually filled in a template (i.e., discrete prompt) to allow the models to understand a given task, but the models usually need extensive fine-tuning to bridge the user/item IDs and the template words and to unleash the power of LLM for recommendation. To address the problems, we propose to distill the discrete prompt for a specific task to a set of continuous prompt vectors so as to bridge IDs and words and to reduce the inference time. We also design a training strategy with an attempt to improve the efficiency of training these models. Experimental results on three real-world datasets demonstrate the effectiveness of our PrOmpt Distillation (POD) approach on both sequential recommendation and top-N recommendation tasks. Although the training efficiency can be significantly improved, the improvement of inference efficiency is limited. This finding may inspire researchers in the community to further improve the inference efficiency of LLM-based recommendation models.

Foundation Models such as Large Language Models (LLMs) have significantly advanced many research areas. In particular, LLMs offer significant advantages for recommender systems, making them valuable tools for personalized recommendations. For example, by formulating various recommendation tasks such as rating prediction, sequential recommendation, straightforward recommendation, and explanation generation into language instructions, LLMs make it possible to build universal recommendation engines that can handle different recommendation tasks. Additionally, LLMs have a remarkable capacity for understanding natural language, enabling them to comprehend user preferences, item descriptions, and contextual information to generate more accurate and relevant recommendations, leading to improved user satisfaction and engagement. This tutorial introduces Foundation Models such as LLMs for recommendation. We will introduce how recommender system advanced from shallow models to deep models and to large models, how LLMs enable generative recommendation in contrast to traditional discriminative recommendation, and how to build LLM-based recommender systems. We will cover multiple perspectives of LLM-based recommendation, including data preparation, model design, model pre-training, fine-tuning and prompting, multi-modality and multi-task learning, as well as trustworthy perspectives of LLM-based recommender systems such as fairness and transparency.

Negative sampling is essential for implicit-feedback-based collaborative filtering, which is used to constitute negative signals from massive unlabeled data to guide supervised learning. The state-of-the-art idea is to utilize hard negative samples that carry more useful information to form a better decision boundary. To balance efficiency and effectiveness, the vast majority of existing methods follow the two-pass approach, in which the first pass samples a fixed number of unobserved items by a simple static distribution and then the second pass selects the final negative items using a more sophisticated negative sampling strategy. However, selecting negative samples from the original items in a dataset is inherently restricted due to the limited available choices, and thus may not be able to contrast positive samples well. In this paper, we confirm this observation via carefully designed experiments and introduce two major limitations of existing solutions: ambiguous trap and information discrimination.

Negative sampling is essential for implicit collaborative filtering to generate negative samples from massive unlabeled data. Unlike existing strategies that consider items as a whole when selecting negative items, we argue that normally user interactions are mainly driven by some relevant, but not all, factors of items, leading to a new direction of negative sampling. In this paper, we introduce a novel disentangled negative sampling (DENS) method. We first disentangle the relevant and irrelevant factors of positive and negative items using a hierarchical gating module. Next, we design a factor-aware sampling strategy to identify the best negative samples by contrasting the relevant factors while keeping irrelevant factors similar. To ensure the credibility of the disentanglement, we propose to adopt contrastive learning and introduce four pairwise contrastive tasks, which enable to learn better disentangled representations of the relevant and irrelevant factors and remove the dependency on ground truth. Extensive experiments on five real-world datasets demonstrate the superiority of DENS against several state-of-the-art competitors, achieving over 7% improvement over the strongest baseline in terms of [email protected] and [email protected] Our code is publically available at https://github.com/Riwei-HEU/DENS .

Online music platforms that include social networking features sometimes become supportive social communities where young people can disclose their emotional distress and receive support. However, few studies have examined young people's disclosure in social music communities or the support they provide or receive. In this study, which focuses on a large online music platform as a research site, we used mixed methods to analyze young users' comments (N = 163) and the associated replies (N = 2,732) related to their psychological distress (e.g., depression, anxiety, stress, and loneliness). We found that the main types of comments involved experience sharing, and these comments often invoked peer support in the form of encouragement, caring, or self-disclosure. We also conducted an interview study with 13 young users of our research site to understand their perceptions of and motives for engaging in disclosure and support. The interviewees stated that music-induced and comment-induced emotional resonance was the main reason for their disclosure and support. Finally, we discussed the implications of our findings for designing a supportive social music community to benefit youth mental health.

Users’ personality traits can take an active role in affecting their behavior when they interact with a computer interface. However, in the area of recommender systems (RS) , though personality-based RS has been extensively studied, most works focus on algorithm design, with little attention paid to studying whether and how the personality may influence users’ interaction with the recommendation interface. In this manuscript, we report the results of a user study (with 108 participants) that not only measured the influence of users’ personality traits on their perception and performance when using the recommendation interface but also employed an eye-tracker to in-depth reveal how personality may influence users’ eye-movement behavior. Moreover, being different from related work that has mainly been conducted in a single product domain, our user study was performed in three typical application domains (i.e., electronics like smartphones, entertainment like movies, and tourism like hotels). Our results show that mainly three personality traits, i.e., Openness to experience , Conscientiousness , and Agreeableness , significantly influence users’ perception and eye-movement behavior, but the exact influences vary across the domains. Finally, we provide a set of guidelines that might be constructive for designing a more effective recommendation interface based on user personality.

Explaining to users why some items are recommended is critical, as it can help users to make better decisions, increase their satisfaction, and gain their trust in recommender systems (RS). However, existing explainable RS usually consider explanation as a side output of the recommendation model, which has two problems: (1) It is difficult to evaluate the produced explanations, because they are usually model-dependent, and (2) as a result, how the explanations impact the recommendation performance is less investigated. In this article, explaining recommendations is formulated as a ranking task and learned from data, similarly to item ranking for recommendation. This makes it possible for standard evaluation of explanations via ranking metrics (e.g., Normalized Discounted Cumulative Gain). Furthermore, this article extends traditional item ranking to an item–explanation joint-ranking formalization to study if purposely selecting explanations could reach certain learning goals, e.g., improving recommendation performance. A great challenge, however, is that the sparsity issue in the user-item-explanation data would be inevitably severer than that in traditional user–item interaction data, since not every user–item pair can be associated with all explanations. To mitigate this issue, this article proposes to perform two sets of matrix factorization by considering the ternary relationship as two groups of binary relationships. Experiments on three large datasets verify the solution’s effectiveness on both explanation ranking and item recommendation.

Providing user-understandable explanations to justify recommendations could help users better understand the recommended items, increase the system’s ease of use, and gain users’ trust. A typical approach to realize it is natural language generation. However, previous works mostly adopt recurrent neural networks to meet the ends, leaving the potentially more effective pre-trained Transformer models under-explored. In fact, user and item IDs, as important identifiers in recommender systems, are inherently in different semantic space as words that pre-trained models were already trained on. Thus, how to effectively fuse IDs into such models becomes a critical issue. Inspired by recent advancement in prompt learning, we come up with two solutions: find alternative words to represent IDs (called discrete prompt learning) and directly input ID vectors to a pre-trained model (termed continuous prompt learning). In the latter case, ID vectors are randomly initialized but the model is trained in advance on large corpora, so they are actually in different learning stages. To bridge the gap, we further propose two training strategies: sequential tuning and recommendation as regularization. Extensive experiments show that our continuous prompt learning approach equipped with the training strategies consistently outperforms strong baselines on three datasets of explainable recommendation.

Recommender systems have become a ubiquitous part of our digital lives.Every day we receive suggestions about what to shop online, which video series to watch, what news to read, or which people to follow on social media.With their suggestions, these systems help us find relevant things when there is just too much information available or when there are too many alternatives to choose from.Over the past 25 years, recommender systems have become a research field of their own, transcending early roots in information retrieval and other areas of computer science.Indeed, building a successful recommender system in practice requires the cooperation of people with different backgrounds, from machine learning and data science experts to human-computer interaction and business professionals.Given the multi-faceted nature of recommender systems, research in this area has traditionally been scattered, with various publication outlets in different scientific subfields.This is why we launched ACM Transactions on Recommender Systems (ACM TORS).With this new journal, our goal is to be a home and preferred outlet for the latest high-quality research on the various aspects of recommender systems, from algorithms to the user experience, to questions of the impact and value of such systems.The journal welcomes submissions of various types, from original research works, to survey papers, reproducibility studies, industry reports, and opinion pieces.This inaugural issue contains a selection of articles on the latest research in our area.Several additional articles have already been accepted for future issues, and we are happy to observe that our research area is flourishing and continuously growing.The idea of a journal entirely devoted to recommender systems has been around for several years.Making this idea a reality would not have been possible without the support of many people.First, there was tremendous support from our research community, both from academia and industry, and from the Steering Committee of the ACM Conference on Recommender Systems.We are thankful that with this support ACM TORS can rely on an extraordinary board of Associate Editors, involving many top-notch international experts who have shaped the entire field over the years.Moreover, many very senior colleagues immediately volunteered to help as part of the board of Distinguished Reviewers, and researchers around the world have already served as reviewers for the articles we have received so far.Our specific thanks go to Joseph A. Konstan, who helped us in several important ways throughout the process, and to Joeran Beel, who is not only the Information Director of ACM TORS but also played an important role when we started our journey.We also thank the people at ACM and their supporting organizations for their help in ramping up this journal.This includes Laura Lander, Shayla Poling, and Starratt Scheetz, who were always there to deal with the questions and inquiries that have come up over the months.Finally, we thank you, the readers of ACM TORS.We are hoping that you will follow the subsequent issues of the journal and that you will enjoy the breadth and variety of the works that we will publish on the various facets of recommender systems research.We are open to your feedback to continuously improve the journal, and we hope that you may consider contributing to the journal in the future.

Conversational recommender systems (CRSs) imitate human advisors to assist users in finding items through conversations and have recently gained increasing attention in domains such as media and e-commerce. Like in human communication, building trust in human-agent communication is essential given its significant influence on user behavior. However, inspiring user trust in CRSs with a “one-size-fits-all” design is difficult, as individual users may have their own expectations for conversational interactions (e.g., who, user or system, takes the initiative), which are potentially related to their personal characteristics. In this study, we investigated the impacts of three personal characteristics, namely personality traits, trust propensity, and domain knowledge, on user trust in two types of text-based CRSs, i.e., user-initiative and mixed-initiative. Our between-subjects user study (N=148) revealed that users’ trust propensity and domain knowledge positively influenced their trust in CRSs, and that users with high conscientiousness tended to trust the mixed-initiative system.

Heterogeneous sequential recommendation (HSR) is a very important recommendation problem, which aims to predict a user’s next interacted item under a target behavior type (e.g., purchase in e-commerce sites) based on his/her historical interactions with different behaviors. Though existing sequential methods have achieved advanced performance by considering the varied impacts of interactions with sequential information, a large body of them still have two major shortcomings. Firstly, they usually model different behaviors separately without considering the correlations between them. The transitions from item to item under diverse behaviors indicate some users’ potential behavior manner. Secondly, though the behavior information contains a user’s fine-grained interests, the insufficient consideration of the local context information limits them from well understanding user intentions. Utilizing the adjacent interactions to better understand a user’s behavior could improve the certainty of prediction. To address these two issues, we propose a novel solution utilizing global and personalized graphs for HSR (GPG4HSR) to learn behavior transitions and user intentions. Specifically, our GPG4HSR consists of two graphs, i.e., a global graph to capture the transitions between different behaviors, and a personalized graph to model items with behaviors by further considering the distinct user intentions of the adjacent contextually relevant nodes. Extensive experiments on four public datasets with the state-of-the-art baselines demonstrate the effectiveness and general applicability of our method GPG4HSR.

Dialogue-based conversational recommender systems (DCRSs) have become a new trend in recommender systems (RSs), allowing users to communicate with the system in natural language to facilitate feedback provision and product exploration. However, little work has been done to empirically study user perception of and interaction with such systems and, more importantly, how to best support users in providing feedback on the recommendation they receive. In this article, we aim to develop effectivecritiquingmechanisms for DCRS to improve its feedback elicitation process (i.e., allowing users tocritiquethe current recommendation during the dialogue). Specifically, we have implemented three prototype systems featuring three different critiquing techniques, respectively, i.e.,user-initiated critiquing, progressive system-suggested critiquing, andcascading system-suggested critiquing. We have then conducted two task-oriented user studies involving 292 subjects to evaluate the three prototypes. In particular, we consider two typical types of user tasks in RSs: basic recommendation task (BRT, i.e., looking for items according to the user’s preferences), and exploration-oriented task (EOT, i.e., exploring different types of items). Results show that EOT stimulates more user interaction, while BRT results in higher user satisfaction. Moreover, when users perform EOT, the type of critiquing techniques is more likely to influence user perception and moderate the relationships between certain interaction metrics and users’ perceived serendipity. The findings suggest effective critiquing techniques to enhance the interaction between users and the recommendation chatbot when the system makes recommendations for different purposes.

Lei Li, Yongfeng Zhang, Li Chen. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

An increasing number of recommender systems enable conversational interaction to enhance the system’s overall user experience (UX). However, it is unclear what qualities of a conversational recommender system (CRS) are essential to determine the success of a CRS. This paper presents a model to capture the key qualities of conversational recommender systems and their related user experience aspects. Our model incorporates the characteristics of conversations (such as adaptability, understanding, response quality, rapport, humanness, etc.) in four major user experience dimensions of the recommender system: User Perceived Qualities, User Belief, User Attitudes, and Behavioral Intentions. Following the psychometric modeling method, we validate the combined metrics using the data collected from an online user study of a conversational music recommender system. The user study results 1) support the consistency, validity, and reliability of the model that identifies seven key qualities of a CRS; and 2) reveal how conversation constructs interact with recommendation constructs to influence the overall user experience of a CRS. We believe that the key qualities identified in the model help practitioners design and evaluate conversational recommender systems.

Dialogue-based conversational recommender systems allow users to give language-based feedback on the recommended item, which has great potential for supporting users to explore the space of recommendations through conversation. In this work, we consider incorporating critiquing techniques into conversational systems to facilitate users’ exploration of music recommendations. Thus, we have developed a music chatbot with three system variants, which are respectively featured with three different critiquing techniques, i.e., user-initiated critiquing (UC), progressive system-suggested critiquing (Progressive SC), and cascading system-suggested critiquing (Cascading SC). We conducted a between-subject study (N=107) to compare these three types of systems with regards to music exploration in terms of user perception and user interaction. Results show that both UC and SC are useful for music exploration, while users perceive higher diversity of recommendations with the system that offers Cascading SC and perceive more serendipitous with the system that offers Progressive SC. In addition, we find that the critiquing techniques significantly moderate the relationships between some interaction metrics (e.g., number of listened songs, number of dialogue turns) and users’ perceived helpfulness and serendipity during music exploration.

Chatbots use conversational interfaces to simulate human communication and recently have been applied to different domains due to advancing techniques of natural language understanding and generation. In particular, in the domain of digital journalism, chatbots provide a new channel for audiences to engage with news. However, most chatbots operated by news organizations have so far failed to achieve business growth. In this paper, we have conducted a qualitative user study to better understand users' attitudes towards news chatbots. Specifically, 15 participants were asked to interact with several news chatbots implemented on Facebook Messenger by large international news organizations (e.g., ABC News, NBC News, and BBC News), by issuing a set of 22 sample questions covering various search and recommendation goals related to COVID-19. Then the participants expressed their expectations of news chatbots and the advantages/disadvantages perceived in using them. From these findings, we derive several design guidelines on effectiveness, informativeness, efficiency, humanization, and facility, suggesting developments to news chatbots that may better serve users' needs.

There are various biases in recommender systems. Recognizing biases, as well as unfairness caused by problematic biases, is the first step of system optimization. Related studies on algorithmic biases are mainly from the perspective of either items or users. For the latter (we call it “algorithmic user bias”), existing works have considered algorithms’ accuracy performances measured by accuracy metrics like RMSE. However, algorithmic user biases in beyond-accuracy measurements have rarely been studied, even though beyond-accuracy oriented recommendation algorithms have been increasingly investigated, with the purpose of breaking through the personalization limits of traditional accuracy-oriented algorithms (such as the typical “filter bubble” phenomenon). To fill in the research gap, in this work, we employ a large-scale survey dataset collected from a commercial platform, in which more than 11,000 users’ ratings on the recommendation’s 5 performance objectives (i.e., relevance, diversity, novelty, unexpectedness, and serendipity) and 8 kinds of user characteristics (i.e., gender, age, big-5 personality traits, and curiosity) are available. We study user biases of four algorithms (i.e., HOT, Rel-CF, Nov-CF, and Ser-CF) in terms of those five measurements between user groups of the eight user characteristics. We further look into users’ behavior patterns like the preference of using more positive ratings, in order to interpret the observed biases. Finally, based on the observed algorithmic user bias and users’ behavior patterns, we analyze the possible factors leading to the biases and recognize problematic biases that may lead to unfairness.

Recently, research on explainable recommender systems has drawn much attention from both academia and industry, resulting in a variety of explainable models. As a consequence, their evaluation approaches vary from model to model, which makes it quite difficult to compare the explainability of different models. To achieve a standard way of evaluating recommendation explanations, we provide three benchmark datasets for EXplanaTion RAnking (denoted as EXTRA), on which explainability can be measured by ranking-oriented metrics. Constructing such datasets, however, poses great challenges. First, user-item-explanation triplet interactions are rare in existing recommender systems, so how to find alternatives becomes a challenge. Our solution is to identify nearly identical sentences from user reviews. This idea then leads to the second challenge, i.e., how to efficiently categorize the sentences in a dataset into different groups, since it has quadratic runtime complexity to estimate the similarity between any two sentences. To mitigate this issue, we provide a more efficient method based on Locality Sensitive Hashing (LSH) that can detect near-duplicates in sub-linear time for a given query. Moreover, we make our code publicly available to allow researchers in the community to create their own datasets.

Personalized recommender systems are important to assist user decision-making in the era of information overload. Meanwhile, explanations of the recommendations further help users to better understand the recommended items so as to make informed choices, which gives rise to the importance of explainable recommendation research. Textual sentence-based explanation has been an important form of explanations for recommender systems due to its advantage in communicating rich information to users. However, current approaches to generating sentence explanations are either limited to predefined sentence templates, which restricts the sentence expressiveness, or opt for free-style sentence generation, which makes it difficult for sentence quality control. In an attempt to benefit both sentence expressiveness and quality, we propose a Neural Template (NETE) explanation generation framework, which brings the best of both worlds by learning sentence templates from data and generating template-controlled sentences that comment about specific features. Experimental results on real-world datasets show that NETE consistently outperforms state-of-the-art explanation generation approaches in terms of sentence quality and expressiveness. Further analysis on case study also shows the advantages of NETE on generating diverse and controllable explanations.

To develop a multi-turn dialogue-based conversational recommender system (DCRS), it is important to predict users' intents behind their utterances and their satisfaction with the recommendation, so as to allow the system to incrementally refine user preference model and adjust its dialogue strategy. However, little work has investigated these issues so far. In this paper, we first contribute with two hierarchical taxonomies for classifying user intents and recommender actions respectively based on grounded theory. We then define various categories of feature considering content, discourse, sentiment, and context to predict users' intents and satisfaction by comparing different machine learning methods. The experimental results for user intent prediction task show that some models (such as XGBoost and SVM) can perform well in predicting user intents, and incorporating context features into the prediction model can significantly boost the performance. Our empirical study also demonstrates that leveraging dialogue behavior features (i.e., including both user intents and recommender actions) can achieve good results in predicting user satisfaction.

Serendipity-oriented recommender systems have increasingly been recognized as useful to overcome the "filter bubble" problem of accuracy-oriented recommenders, by recommending unexpected and relevant items to users. However, most of existing systems are based on researchers' assumptions about the effect of item features on serendipity, but less from users' perspective to study what item features and even user characteristics might affect their perceived serendipity. In this paper, we have attempted to fill in this vacancy based on results of a large-scale user survey (involving over 10,000 users). We have analyzed the correlation between different types of features (i.e., numerical and categorical) with user perceptions, and furthermore identified the interaction effect from user characteristics (such as personality traits and curiosity). We finally discuss the implications of our work to augment the effectiveness of current serendipity-oriented recommender systems.

Critiquing-based recommender systems aim to elicit more accurate user preferences from users' feedback toward recommendations. However, systems using a graphical user interface (GUI) limit the way that users can critique the recommendation. With the rise of chatbots in many application domains, they have been regarded as an ideal platform to build critiquing-based recommender systems. Therefore, we present MusicBot, a chatbot for music recommendations, featured with two typical critiquing techniques, user-initiated critiquing (UC) and system-suggested critiquing (SC). By conducting a within-subjects (N=45) study with two typical scenarios of music listening, we compared a system of only having UC with a hybrid critiquing system that combines SC with UC. Furthermore, we analyzed the effects of four personal characteristics,musical sophistication (MS), desire for control (DFC), chatbot experience (CE), and tech savviness (TS), on the user's perception and interaction of the recommendation in MusicBot. In general, compared with UC, SC yields higher perceived diversity and efficiency in looking for songs; combining UC and SC tends to increase user engagement. Both MS and DFC positively influence several key user experience (UX) metrics of MusicBot such as interest matching, perceived controllability, and intent to provide feedback.

Recommendation serendipity is being increasingly recognized as being equally important as the other beyond-accuracy objectives (such as novelty and diversity), in eliminating the “filter bubble” phenomenon of the traditional recommender systems. However, little work has empirically verified the effects of serendipity on increasing user satisfaction and behavioral intention. In this paper, we report the results of a large-scale user survey (involving over 3,000 users) conducted in an industrial mobile e-commerce setting. The study has identified the significant causal relationships from novelty, unexpectedness, relevance, and timeliness to serendipity, and from serendipity to user satisfaction and purchase intention. Moreover, our findings reveal that user curiosity plays a moderating role in strengthening the relationships from novelty to serendipity and from serendipity to satisfaction. Our third contribution lies in the comparison of several recommender algorithms, which demonstrates the significant improvements of the serendipity-oriented algorithm over the relevance- and novelty-oriented approaches in terms of user perceptions. We finally discuss the implications of this experiment, which include the feasibility of developing a more precise metric for measuring recommendation serendipity, and the potential benefit of a curiosity-based personalized serendipity strategy for recommender systems.

The explanation interface has been recognized as important in recommender systems because it can allow users to better judge the relevance of recommendations to their preferences and, hence, make more informed decisions. In different product domains, the specific purpose of explanation can be different. For high-investment products (e.g., digital cameras, laptops), how to educate the typical type of new buyers about product knowledge and, consequently, improve their preference certainty and decision quality is essentially crucial. With this objective, we have developed a novel tradeoff-oriented explanation interface that particularly takes into account sentiment features as extracted from product reviews to generate recommendations and explanations in a category structure. In this manuscript, we first reported the results of an earlier user study (in both before-after and counter-balancing setups) that compared our prototype system with the traditional one that purely considers static specifications for explanations. This experiment revealed that adding sentiment-based explanations can significantly increase users’ product knowledge, preference certainty, perceived information usefulness, perceived recommendation transparency and quality, and purchase intention. In order to further identify the reason behind users’ perception improvements on the sentiment-based explanation interface, we performed a follow-up lab controlled eye-tracking experiment that investigated how users viewed information and compared products on the interface. This study shows that incorporating sentiment features into the tradeoff-oriented explanations can significantly affect users’ eye-gaze pattern. They were stimulated to not only notice bottom categories of products, but also, more frequently, to compare products across categories. The results also disclose users’ inherent information needs for sentiment-based explanations, as they allow users to better understand the recommended products and gain more knowledge about static specifications.

The explanation interface has been recognized important in recommender systems as it can help users evaluate recommendations in a more informed way for deciding which ones are relevant to their interests. In different decision environments, the specific aim of explanation can be different. In high-investment product domains (e.g., digital cameras, laptops) for which users usually attempt to avoid financial risk, how to support users to construct stable preferences and make better decisions is particularly crucial. In this paper, we propose a novel explanation interface that emphasizes explaining the tradeoff properties within a set of recommendations in terms of both their static specifications and feature sentiments extracted from product reviews. The objective is to assist users in more effectively exploring and understanding product space, and being able to better formulate their preferences for products by learning from other customers' experiences. Through two user studies (in form of both before-after and within-subjects experiments), we empirically identify the practical role of feature sentiments in combination with static specifications in producing tradeoff-oriented explanations. Specifically, we find that our explanation interface can be more effective to increase users' product knowledge, preference certainty, perceived information usefulness, recommendation transparency and quality, and purchase intention.

Sense of community is regarded as the reflection of students' feelings of connectedness with community members and commonality of learning expectations and goals. In online courses, sense of community has been proven to influence students' learning engagement and academic performance. Low sense of community is also one of the reasons for drop out. However, existing studies mainly acquire students' sense of community via questionnaires, which demand user efforts and have difficulty in obtaining real-time feeling during students' learning process. In addition, although communication is helpful to enhance students' sense of community, little work has empirically compared the impact of different online communication tools. In this paper, we are motivated to derive students' sense of community from their communication behavior in online courses. Concretely, we first identify a set of features that are significantly correlated with students' sense of community, which not only include their activities carried out in both synchronous and asynchronous online learning environment, but also their linguistic content in conversational texts. We then develop inference model to unify these features for determining students' sense of community, and find that LASSO performs the best in terms of inference accuracy.

When opening a new restaurant, geographical placement is of prime importance in determining whether it will thrive. Although some methods have been developed to assess the attractiveness of candidate locations for a restaurant, the accuracy is limited as they mainly rely on traditional data sources, such as demographic studies or consumer surveys. With the advent of abundant user-generated restaurant reviews, there is a potential to leverage these reviews to gain some insights into users' preferences for restaurants. In this paper, we particularly take advantage of user-generated reviews to construct predictive features for assessing the attractiveness of candidate locations to expand a restaurant. Specifically, we investigate three types of features: review-based market attractiveness, review-based market competitiveness and geographic characteristics of a location under consideration for a prospective restaurant. We devise the three sets of features and incorporate them into a regression model to predict the number of check-ins that a prospective restaurant at a candidate location would be likely to attract. We then conduct an experiment with real-world restaurant data, which demonstrates the predictive power of features we constructed in this paper. Moreover, our experimental results suggest that market attractiveness and market competitiveness features mined solely from user-generated restaurant reviews are more predictive than geographic features.

The critiquing-based recommender system (CBRS) stimulates users to critique the recommended item in terms of its attribute values. It has been shown that such critiquing feedback can effectively improve users' decision quality, especially in complex decision environments such as e-commerce, tourism, and finance. However, because its explicit elicitation process unavoidably demands extra user efforts, the application in real situations is limited. In this paper, we report an eye-tracking experiment with the objective of studying the relationship between users' eye gazes as laid on recommended items and their critiquing feedback. The results indicate the feasibility of inferring users' feedback based on their eye movements. It hence points out a promising roadmap to developing unobtrusive eye-based feedback elicitation for recommender systems.

We present several interrelated technical and empirical contributions to the problem of emotion-based music recommendation and show how they can be applied in a possible usage scenario. The contributions are (1) a new three-dimensional resonance-arousal-valence model for the representation of emotion expressed in music, together with methods for automatically classifying a piece of music in terms of this model, using robust regression methods applied to musical/acoustic features; (2) methods for predicting a listener’s emotional state on the assumption that the emotional state has been determined entirely by a sequence of pieces of music recently listened to, using conditional random fields and taking into account the decay of emotion intensity over time; and (3) a method for selecting a ranked list of pieces of music that match a particular emotional state, using a minimization iteration method. A series of experiments yield information about the validity of our operationalizations of these contributions. Throughout the article, we refer to an illustrative usage scenario in which all of these contributions can be exploited, where it is assumed that (1) a listener’s emotional state is being determined entirely by the music that he or she has been listening to and (2) the listener wants to hear additional music that matches his or her current emotional state. The contributions are intended to be useful in a variety of other scenarios as well.

Fine-grained opinions are often buried in user reviews. The opinionated aspects may also be associated with different weights by reviewers to represent the aspects’ relative importance. As the opinions and weights provide valuable information about users’ preferences for products, they can facilita te the generation of personalised recommendations. However, few studies to date have investigated the three inter-connected tasks in a unified framework: aspect identification, aspect-based rating inference and weight estimation. In this paper, we propose a unified framework for performing the three tasks, which involves 1) identifying the product aspects mentioned in a review, 2) inferring the reviewer’s ratings for these aspects from the opinions s/he expressed in a review, and 3) estimating the reviewer’s weights for these aspects. The relationship among these three tasks is inherently dependent in that the output of one task adjusts the accuracy of another task. We particularly develop an unsupervised model to Collectively estimate Aspect Ratings and Weights (shorted as CARW), which performs all of the three tasks by enhancing each other mutually. We conduct experiments on three real-life datasets to evaluate the CARW model. Experimental results show that the proposed model can achieve better performance than the related methods regarding each task.

Recommender systems have been increasingly adopted as personalisation services in e-commerce. They facilitate users to locate items which they would be interested in viewing or purchasing. However, most studies have emphasised on the algorithm's performance, rather than on in-depth analysis of user experiences with the recommender interface. In this article, we report the results of two studies that compared two recommender interfaces: the organisation-based interface (where recommendations are presented in a category structure via the preference-based organisation method) and the standard ranked list (where recommendations are listed one after the other as ordered by their prediction scores).The first study focuses on evaluating users' eye-movement behaviour in these interfaces. With the help of an eye tracker, we found that the organisation interface (ORG) can significantly attract users' attentions to more recommended items. As a result, more users made product choices in that interface. The second, larger-scale, cross-cultural user survey further shows that the ORG performed significantly better in terms of enhancing users' perceived recommendation quality, perceived ease of use and perceived usefulness of the system. Hence, these empirical findings suggest that the change of recommender interface design can not only alter users' attention distribution, but also influence their subjective attitudes towards the system.

A primary function of recommender systems is to help their users to make better choices and decisions. The overall goal of the workshop is to analyse and discuss novel techniques and approaches for supporting effective and efficient human decision making in different types of recommendation scenarios. The submitted papers discuss a wide range of topics from core algorithmic issues to the management of the human computer interaction.

Collaborative filtering aims to make use of users’ feedbacks to improve the recommendation performance, which has been deployed in various industry recommender systems. Some recent works have switched from exploiting explicit feedbacks of numerical ratings to implicit feedbacks like browsing and shopping records, since such data are more abundant and easier to collect. One fundamental challenge of leveraging implicit feedbacks is the lack of negative feedbacks, because there are only some observed relatively “positive” feedbacks, making it difficult to learn a prediction model. Previous works address this challenge via proposing some pointwise or pairwise preference assumptions on items. However, such assumptions with respect to items may not always hold, for example, a user may dislike a bought item or like an item not bought yet. In this paper, we propose a new and relaxed assumption of pairwise preferences over item-sets, which defines a user's preference on a set of items (item-set) instead of on a single item. The relaxed assumption can give us more accurate pairwise preference relationships. With this assumption, we further develop a general algorithm called CoFiSet (collaborative filtering via learning pairwise preferences over item-sets). Experimental results show that CoFiSet performs better than several state-of-the-art methods on various ranking-oriented evaluation metrics on two real-world data sets. Furthermore, CoFiSet is very efficient as shown by both the time complexity and CPU time.

Recommender systems have already proved to be valuable for coping with the information overload problem in several application domains. They provide people with suggestions for items which are likely to be of interest for them; hence, a primary function of recommender systems is to help people make good choices and decisions. However, most previous research has focused on recommendation techniques and algorithms, and less attention has been devoted to the decision making processes adopted by the users and possibly supported by the system. There is still a gap between the importance that the community gives to the assessment of recommendation algorithms and the current range of ongoing research activities concerning human decision making. Different decision-psychological phenomena can influence the decision making of users of recommender systems, and research along these lines is becoming increasingly important and popular. This special issue highlights how the coupling of recommendation algorithms with the understanding of human choice and decision making theory has the potential to benefit research and practice on recommender systems and to enable users to achieve a good balance between decision accuracy and decision effort.

No abstract available.

Collaborative filtering (CF) recommenders based on User-Item rating matrix as explicitly obtained from end users have recently appeared promising in recommender systems. However, User-Item rating matrix is not always available or very sparse in some web applications, which has critical impact to the application of CF recommenders. In this article we aim to enhance the online recommender system by fusing virtual ratings as derived from user reviews. Specifically, taking into account of Chinese reviews' characteristics, we propose to fuse the self-supervised emotion-integrated sentiment classification results into CF recommenders, by which the User-Item Rating Matrix can be inferred by decomposing item reviews that users gave to the items. The main advantage of this approach is that it can extend CF recommenders to some web applications without user rating information. In the experiments, we have first identified the self-supervised sentiment classification's higher precision and recall by comparing it with traditional classification methods. Furthermore, the classification results, as behaving as virtual ratings, were incorporated into both user-based and item-based CF algorithms. We have also conducted an experiment to evaluate the proximity between the virtual and real ratings and clarified the effectiveness of the virtual ratings. The experimental results demonstrated the significant impact of virtual ratings on increasing system's recommendation accuracy in different data conditions (i.e., conditions with real ratings and without).

Most products in e-commerce are with high cost (e.g., digital cameras, computers) and hence less likely experienced by users (so they are called "inexperienced products"). The traditional recommender techniques (such as user-based collaborative filtering and content-based methods) are thus not effectively applicable in this environment, because they largely assume that the users have prior experiences with the items. In this paper, we have particularly incorporated product reviews to solve the recommendation problem. We first studied how to utilize the reviewer-level weighted feature preferences (as learnt from their written product reviews) to generate recommendations to the current buyer, followed by exploring the impact of Latent Class Regression Models (LCRM) based cluster-level feature preferences (that represent the common preferences of a group of reviewers). Motivated by their respective advantages, a hybrid method that combines both reviewer-level and cluster-level preferences is introduced and experimentally compared to the other methods. The results reveal that the hybrid method is superior to the other variations in terms of recommendation accuracy, especially when the current buyer states incomplete feature preferences.

In this paper, we report the hotspot and gaze path of users' eye-movements on three different layouts for recommender interfaces. One is the standard list layout, as appearing in most of current recommender systems. The other two are variations of organization interfaces where recommended items are organized into categories and each category is annotated by a title. Gaze plots infer that the organization interfaces, especially the quadrant layout, are likely to arouse users' attentions to more recommendations. In addition, more users chose products from the organization layouts. Combining the results with our prior works, we suggest a set of design guidelines and practical implications to our future work.

Interacting with a recommender system means to take different decisions such as selecting a song/movie from a recommendation list, selecting specific feature values (e.g., camera's size, zoom) as criteria, selecting feedback features to be critiqued in a critiquing based recommendation session, or selecting a repair proposal for inconsistent user preferences when interacting with a knowledge-based recommender. In all these scenarios, users have to solve a decision task. The major focuses of this workshop ([email protected]) were approaches for efficient human decision making in different types of recommendation scenarios.

This research was motivated by our interest in understanding the criteria for measuring the success of a recommender system from users' point view. Even though existing work has suggested a wide range of criteria, the consistency and validity of the combined criteria have not been tested. In this paper, we describe a unifying evaluation framework, called ResQue (Recommender systems' Quality of user experience), which aimed at measuring the qualities of the recommended items, the system's usability, usefulness, interface and interaction qualities, users' satisfaction with the systems, and the influence of these qualities on users' behavioral intentions, including their intention to purchase the products recommended to them and return to the system. We also show the results of applying psychometric methods to validate the combined criteria using data collected from a large user survey. The outcomes of the validation are able to 1) support the consistency, validity and reliability of the selected criteria; and 2) explain the quality of user experience and the key determinants motivating users to adopt the recommender technology. The final model consists of thirty two questions and fifteen constructs, defining the essential qualities of an effective and satisfying recommender system, as well as providing practitioners and scholars with a cost-effective way to evaluate the success of a recommender system and identify important areas in which to invest development resources.

Collaborative Filtering (CF) based recommender systems often suffer from the sparsity problem, particularly for new and inactive users when they use the system. The emerging trend of social networking sites and their accommodation in other sites like e-commerce can potentially help alleviate the sparsity problem with their provided social relation data. In this paper, we have particularly explored a new kind of social relation, the membership, and its combined effect with friendship. The two type of heterogeneous social relations are fused into the CF recommender via a factorization process. Due to the two relations' respective properties, we adopt different fusion strategies: regularization was leveraged for friendship and collective matrix factorization (CMF) was proposed for incorporating membership. We further developed a unified model to combine the two relations together and tested it with real large-scale datasets at five sparsity levels. The experiment has not only revealed the significant effect of the two relations, especially the membership, in augmenting recommendation accuracy in the sparse data condition, but also identified the ability of our fusing model in achieving the desired fusion performance.

The tasks of feature-level opinion mining usually include the extraction of product entities from product reviews, the identification of opinion words that are associated with the entities, and the determining of these opinions' polarities (e.g., positive, negative, or neutral). In recent years, several approaches have been proposed such as rule-based and statistical methods on this subject, but few attentions have been paid to applying more discriminative learning models to achieve the goal. On the other hand, little work has evaluated their algorithms' performance for identifying intensifiers, entity phrases and infrequent entities. In this paper, we in particular adopt the Conditional Random Fields (CRFs) model to perform the opinion mining tasks. Relative to related approaches, we have not only highlighted the algorithm's ability in mining intensifiers, phrases and infrequent entities, but also integrated more elements in the model so as to optimize its training and decoding process. Our method was compared to the lexicalized Hidden Markov Model (L-HMMs) based opinion mining method in the experiment, which proves its significantly better accuracy from several aspects.

No abstract available.

This paper aims at studying the roles of social information (as obtained from social networking sources) in buyers' decision process when they are searching for products to buy. Through both close observation of users' objective behavior and qualitative interview of their reflective comments, we have discovered the importance of different types of social information in supporting users to achieve a confident purchase decision at the end. Improving suggestions are further derived on how to better present the social information and combine them with static product attributes to enhance current e-commerce decision supports so as to optimally adapt to online buyers' actual information needs.

"Product popularity" is in-depth explored in this paper, regarding its practical role within a consumer's decision process. Specifically, the usability evaluation of a novel product finder service (Flickr Camera Finder) shows that users more frequently consulted it, rather than a standard shopping site, to locate popular products. User comments further revealed their credibility concerns and tendency to trust the "popularity" from social resources. Design implications from the experiment are summarized at the end, indicating suggestive directions to integrate social media data to boost current e-commerce decision tools.

This workshop brought researchers from academia and industry together to share recent advances and discuss research directions for recommender systems in social media and Web 2.0. With social media sites becoming ubiquitous, the challenges and opportunities for recommendation technologies become greater, setting the grounds for new research and innovation.

Accurately capturing user preferences over time is a great practical challenge in recommender systems. Simple correlation over time is typically not meaningful, since users change their preferences due to different external events. User behavior can often be determined by individual's long-term and short-term preferences. How to represent users' long-term and short-term preferences? How to leverage them for temporal recommendation? To address these challenges, we propose Session-based Temporal Graph (STG) which simultaneously models users' long-term and short-term preferences over time. Based on the STG model framework, we propose a novel recommendation algorithm Injected Preference Fusion (IPF) and extend the personalized Random Walk for temporal recommendation. Finally, we evaluate the effectiveness of our method using two real datasets on citations and social bookmarking, in which our proposed method IPF gives 15%-34% improvement over the previous state-of-the-art.

As e-commerce has evolved into its second generation, where the available products are becoming more complex and their abundance is almost unlimited , the task of locating a desired choice has become too difficult for the average user. Therefore, more effort has been made in recent years to develop recommender systems that recommend products or services to users so as to assist in their decision-making process. In this article, we describe crucial experimental results about a novel recommender technology, called the preference-based organization (Pref-ORG), which generates critique suggestions in addition to recommendations according to users' preferences. The critique is a form of feedback (“I would like something cheaper than this one”) that users can provide to the currently displayed product, with which the system may better predict what the user truly wants. We compare the preference-based organization technique with related approaches, including the ones that also produce critique candidates, but without the consideration of user preferences. A simulation setup is first presented, that identified Pref-ORG's significantly higher algorithm accuracy in predicting critiques and choices that users should intend to make, followed by a real-user evaluation which practically verified its significant impact on saving users' decision effort.

The completeness and certainty of a user's preferences may vary during her preference construction process in a conversational recommender. In order to more effectively support users to uncover their hidden criteria and/or solve preference conflicts, we propose to generate adaptive tradeoff explanations in organization-based recommender interfaces, to be conditional on the user's contextual needs. An experiment shows the adaptive element's higher potential to improve recommendation efficiency, relative to methods without this feature.

We present a cross-cultural user evaluation of an organization-based product recommender interface, by comparing it with the traditional list view. The results show that it performed significantly better, for all study participants, in improving on their competence perceptions, including perceived recommendation quality, perceived ease of use and perceived usefulness, and positively impacting users' behavioral intentions such as intention to save effort in the next visit. Additionally, oriental users were observed reacting more significantly strongly to the organization interface regarding some subjective aspects, compared to western subjects. Through this user study, we also identified the dominating role of the recommender system's decision-aiding competence in stimulating both oriental and western users' return intention to an e-commerce website where the system is applied.

We propose a novel critiquing-based recommender interface, the hybrid critiquing interface that integrates the user self-motivated critiquing facility to compensate for the limitations of system-proposed critiques. The results from our user study show that the integration of such self-motivated critiquing support enables users to achieve a higher level of decision accuracy while consuming less cognitive effort. In addition, users expressed higher subjective opinions of the hybrid critiquing interface than the interface simply providing system-proposed critiques, and they would more likely return to it for future use.

The critiquing-based recommender system mainly aims to guide users to make an accurate and confident decision, while requiring them to consume a low level of effort. We have previously found that the hybrid critiquing system of combining the strengths from both system-proposed critiques and user self-motivated critiquing facility can highly improve users' subjective perceptions such as their decision confidence and trusting intentions. In this paper, we continue to investigate how to further reduce users' objective decision effort (e.g. time consumption) in such system by increasing the critique prediction accuracy of the system-proposed critiques. By means of real user evaluation, we proved that a new hybrid critiquing system design that integrates the preference-based recommendations organization technique for critiques suggestion can effectively help to increase the proposed critiques' application frequency and significantly contribute to saving users' task time and interaction effort.

Based on our recent work on the development of a trust model for recommender agents and a qualitative survey, we explore the potential of building users' trust with explanation interfaces. We present the major results from the survey, which provided a roadmap identifying the most promising areas for investigating design issues for trust-inducing interfaces. We then describe a set of general principles derived from an in-depth examination of various design dimensions for constructing explanation interfaces, which most contribute to trust formation. We present results of a significant-scale user study, which indicate that the organization-based explanation is highly effective in building users' trust in the recommendation interface, with the benefit of increasing users' intention to return to the agent and save cognitive effort.

In a previously reported user study, we found that users were able to perform decision tradeoff tasks more efficiently and commit considerably fewer errors with the example critiquing interface than with the ranked list. We concluded that example-based search tools were likely to be useful particularly for extending the scope of consumer e-commerce to more complex products where decision making is critical. This paper presents results from a follow-up user study quantifying the benefits of tradeoff support. Users were able to refine the quality of their preference structures and improve decision accuracy by up to 57% after performing tradeoff tasks. Tradeoff support also significantly increased users' confidence in their choices. Together, these two studies show that example critiquing enables users to more accurately find what they want and be confident in their choices, while only requiring a level of effort that is comparable to the ranked list interface.