Nowadays, many online services use recommendation systems to provide personalized item recommendations to users. Collaborative filtering is the major paradigm in recommendation systems. Based on user-item interaction data, collaborative filtering recommends items to a user based on other similar users. The problem of interest disentanglement in recommendation now has attracted the attention of many researchers. Several works have proposed methods to disentangle conformity from user private interest, by assuming that conformity is correlated to item popularity. However, such modeling is simplistic and overlooks many possibilities between user public and private interest, and the item popularity. For example, a user can privately like a popular movie or buy a niche music album due to the stimulation of the social environment. In this paper, we propose a more comprehensive social behavior model that describes fine-grained relationships between user interest and item popularity. Our model does not use explicit user relationship data. Instead, we extract social behavior patterns directly from user-item interaction data. We also make our model into a recommendation framework called Disentangled Social Consumer Preference (DSCP), which can be integrated into existing recommendation models such as BPRMF. Our extensive experiments with four datasets from different services show that our model can outperform state-of-the-art baseline models. We achieve better recommendation accuracy in both the usual random test and the intervened test that shows debiasing effect.

Abstract With the explosive growth of online information, users may also face information overload. To handle this problem, recommender systems have become an effective strategy, which can analyze the characters of users and items to provide valuable information. One of the important types of information is the item’s side information. For example, in Amazon dataset, side information mainly includes visual side information (e.g., image and video), textual side information (e.g., title and description), and auxiliary side information (e.g., brand and category). To analyze various types of side information, some research designed multiple modalities for different types of side information, which can improve the performance of the recommender system. To analyze the deeper relationships between users and items, recent works also use a graph structure to represent the interactions. Existing works on multi-modal recommender systems using graph neural networks largely depend on the interaction records, while little effort focuses on the relationships between interactions and various types of side information. In this paper, we propose a novel multi-task learning model. First, we construct the interaction records to graphs for each modality to gather the representations, and then we analyze the representations of each modality and the specific side information based on the similarities. We design a multi-task multi-modal graph neural network framework built upon message passing with the attention mechanism of graph neural networks, which can generate the representations of users and items from interaction records, and then analyze the relationships between the representations from GNNs and item’s side information. We conduct experiments on three public datasets, Amazon, Modcloth and MovieLens. The results of our model outperform the state-of-the-art methods.

In democratic countries, political interest is deeply involved in people’s daily lives. Research in political consumerism shows that product purchase decision is also influenced by the political orientation of the consumer. In traditional recommendation system design, user interest in an item is provided by a unified model. Recently, interest disentanglement methods have been introduced. It is shown that by disentangling interest factors such as conformity and private interest, recommendation performance can be significantly improved. However, few studies attempt to disentangle political interest in purchase behavior, which is bipolar. In this article, we propose a method to extract political interest model from e-commerce interaction data, which is supported by a novel word-level political bias assignment. For the bias assignment part, we improved a political bias distilling method. For the political interest model extraction part, we extend a one-side bias method to make it support bipolar bias. We compare our method with state-of-the-art baseline methods in several evaluation settings, and the experimental results show that our method can achieve superior performance. Further investigation shows that our method is consistent with theories of political consumerism.

Political polarization is commonly observed in democratic countries. While it allows individual citizens to freely choose sides, it also causes the problem of separation and isolation. Especially in information-seeking behaviors, echo chambers and filter bubbles are observed. In this article, we present a political sentiment dictionary for analyzing political polarizatio and increasing information heterogeneity. It takes advantage of large-scale social media data and is thus superior in accuracy and coverage compared to manually crafted dictionaries. Generated from Japanese tweets, more than 50k words in this dictionary cover aspects ranging from political parties and public entities to foods and personal hobbies. We describe in detail the method to construct this dictionary, which can be replicated for other languages and countries. We demonstrate the use of this dictionary in the application of recommendation diversification. We show with real-world e-commerce data that the use of the dictionary can generally increase the diversity in product recommendations, effectively mitigating the filter bubbles.

Building high-quality recommender systems is challenging for new services and small companies, because of their sparse interactions. Cross-domain recommendations (CDRs) alleviate this issue by transferring knowledge from data in external domains. However, most existing CDRs leverage data from only a single external domain and serve only two domains. CDRs serving multiple domains require domain-shared entities (i.e., users and items) to transfer knowledge, which significantly limits their applications due to the hardness and privacy concerns of finding such entities. We therefore focus on a more general scenario, non-overlapped multi-target CDRs (NO-MTCDRs), which require no domain-shared entities and serve multiple domains. Existing methods require domain-shared users to learn user preferences and cannot work on NO-MTCDRs. We hence propose MITrans, a novel mutual information-based (MI-based) preference disentangling and transferring framework to improve recommendations for all domains. MITrans effectively leverages knowledge from multiple domains as well as learning both domain-shared and domain-specific preferences without using domain-shared users. In MITrans, we devise two novel MI constraints to disentangle domain-shared and domain-specific preferences. Moreover, we introduce a module that fuses domain-shared preferences in different domains and combines them with domain-specific preferences to improve recommendations. Our experimental results on two real-world datasets demonstrate the superiority of MITrans in terms of recommendation quality and application range against state-of-the-art overlapped and non-overlapped CDRs.

Abstract Organized event is an important form of human activity. Nowadays, many digital platforms offer organized events on the Internet, allowing users to be organizers or participants. For such platforms, it is beneficial to predict potential event participants. Existing work on this problem tends to borrow recommendation techniques. However, compared to e-commerce items and purchases, events and participation are usually of a much smaller frequency, and the data may be insufficient to learn an accurate prediction model. In this paper, we propose to utilize social media retweeting activity to enhance the learning of event participant prediction models. We create a joint knowledge graph to bridge the social media and the target domain, assuming that event descriptions and tweets are written in the same language. Furthermore, we propose a learning model that utilize retweeting information for the target domain prediction more effectively. We conduct comprehensive experiments in two scenarios with real-world data. In each scenario, we set up training data of different sizes, as well as warm and cold test cases. The evaluation results show that our approach consistently outperforms several baseline models in both warm and cold tests.

Neural Processes (NPs) are a family of supervised density estimators devoted to probabilistic function approximation with meta-learning. Despite extensive research on the subject, the absence of a unified framework for NPs leads to varied architectural solutions across diverse studies. This non-consensus poses challenges to reproducing and benchmarking different NPs. Moreover, existing codebases mainly prioritize generative density estimation, yet rarely consider expanding the capability of NPs to self-supervised representation learning, which however has gained growing importance in data mining applications. To this end, we present NP-SSL, a modular and configurable framework with built-in support, requiring minimal effort to 1) implement classical NPs architectures; 2) customize specific components; 3) integrate hybrid training scheme (e.g., contrastive); and 4) extend NPs to act as a self-supervised learning toolkit, producing latent representations of data, and facilitating diverse downstream predictive tasks. To illustrate, we discuss a case study that applies NP-SSL to model time-series data. We interpret that NP-SSL can handle different predictive tasks such as imputation and forecasting, by a simple switch in data samplings, without significant change to the underlying structure. We hope this study can reduce the workload of future research on leveraging NPs to tackle more a broader range of real-world data mining applications. Code and documentation are at https://github.com/zyecs/NP-SSL.

Recommender systems nowadays are commonly used in various platforms to provide information based on user preferences. In POI recommendation, systems can generally learn the users' short-term and long-term preferences, which are based on sessions and global information. Existing systems, however, usually overlook the mid-term information, which may contain important indications of user preferences. In this work, we propose a session-based POI recommender system based on Graph Neural Network (GNN). In contrast to existing work, our model can learn short-term, long-term, and mid-term preferences at the same time. In order to learn the mid-term item representation, we construct a week graph and process it by a GAT-based graph model. We further use a gate fusion to integrate three temporal dimensions to obtain the hybrid item representation. We conduct experiments with a real-world POI visiting dataset, and the evaluation results show that our model outperforms compared state-of-art models. By adding mid-term information, the prediction accuracy can be improved by 5% compared to the best baseline.

Recommendation models are typically trained on observational user interaction data, but the interactions between latent factors in users' decision-making processes lead to complex and entangled data. Disentangling these latent factors to uncover their underlying representation can improve the robustness, interpretability, and controllability of recommendation models. This paper introduces the Causal Disentangled Variational Auto-Encoder (CaD-VAE), a novel approach for learning causal disentangled representations from interaction data in recommender systems. The CaD-VAE method considers the causal relationships between semantically related factors in real-world recommendation scenarios, rather than enforcing independence as in existing disentanglement methods. The approach utilizes structural causal models to generate causal representations that describe the causal relationship between latent factors. The results demonstrate that CaD-VAE outperforms existing methods, offering a promising solution for disentangling complex user behavior data in recommendation systems.

Social media as a background information source has been utilized in many practical computational tasks, such as stock price prediction, epidemic tracking, and product recommendation. However, proper representation of an evolving social media background is still in an early research stage. In this article, we propose a representation method that considers temporal novelties as well as the fine details of word inter-dependencies. Our method is based on the tf-idf and graph embedding techniques. The proposed method has superiority over other representation methods because it takes the advantage of both the temporal aspect of tf-idf and the semantic aspect of graph embeddings. We compare our method with a variety of baselines in two practical application scenarios using real-world data. In tweet popularity prediction, our representation achieves 5.7% less error and 12.8% higher correlation compared to the best baseline. In e-commerce product recommendation, our representation achieves 17% higher hit-rate and 20% higher NDCG compared to the best baseline.

Recommender systems nowadays are commonly deployed in e-commerce platforms to help customers making purchase decisions. Dynamic recommender considers not only static user-item interaction data, but the temporal information at the time of recommendation. Previous researches have suggested to incorporate social media as the temporal information in dynamic neural recommenders after transforming them into embeddings. While such an approach can potentially improve recommendation performance, the effectiveness is difficult to explain. In this article, we propose an explainable method to integrate social media in a dynamic neural recommender. Our method applies association rule mining, which can generate human-understandable behavior patterns from social media and e-commerce platforms. With real-world social media and e-commerce data, we show that the integration can improve accuracy by up to 14% while using the same data. Moreover, we can explain the positive cases by examining relevant association rules.

Deep learning-based recommender systems may lead to over-fitting when lacking training interaction data. This over-fitting significantly degrades recommendation performances. To address this data sparsity problem, cross-domain recommender systems (CDRSs) exploit the data from an auxiliary source domain to facilitate the recommendation on the sparse target domain. Most existing CDRSs rely on overlapping users or items to connect domains and transfer knowledge. However, matching users is an arduous task and may involve privacy issues when data comes from different companies, resulting in a limited application for the above CDRSs. Some studies develop CDRSs that require no overlapping users and items by transferring learned user interaction patterns. However, they ignore the bias in user interaction patterns between domains and hence suffer from an inferior performance compared with single-domain recommender systems. In this paper, based on the above findings, we propose a novel CDRS, namely semantic clustering enhanced debiasing graph neural recommender system (SCDGN), that requires no overlapping users and items and can handle the domain bias. More precisely, SCDGN semantically clusters items from both domains and constructs a cross-domain bipartite graph generated from item clusters and users. Then, the knowledge is transferred via this cross-domain user-cluster graph from source to the target. Furthermore, we design a debiasing graph convolutional layer for SCDGN to extract unbiased structural knowledge from the cross-domain user-cluster graph. Our Experimental results on three public datasets and a pair of proprietary datasets verify the effectiveness of SCDGN over stateof-the-art models in terms of cross-domain recommendations.

Nowadays routing systems can provide optimal routes in terms of time and travel distance. However, they do not consider special needs of certain group of users. For example, people recovering from alcohol and drug addiction may want to travel a route that is alcohol and drug-free. In this demonstration, we propose a system we built that helps with this special need. We detect if a street is related to alcohol and drug by exploiting Web open data, including Foursquare, microblog tweets, Google Street View images, and crime data. We calculate an alcohol and drug relevance score using unsupervised methods, to be used in route ranking. Our system prototype is ready to be tested for the cities of San Francisco and Kyoto.

The vast and growing amount of publicly available real-time information from social network services such as Twitter could provide many benefits for improving public health and safety, especially towards the area of crime prevention. While prior studies have leveraged such data to help in the prediction of criminal incidents, we have developed a crime investigation tool which utilizes Twitter data to aid in crime analysis. The tool provides contextual information about crime incidents by visualizing the spatial and time-based characteristics of a crime and its context using data from nearby tweets and from the criminal history of a target place. In addition, sentiment analysis is also carried out with the identified tweets to further examine the negative characteristics of the spatial areas related to the different crimes in question. A demonstration prototype of this tool was developed as a web application for the area of San Francisco.

For a tourist who wishes to stroll in an unknown city, it is useful to have a recommendation of not just the shortest routes but also routes that are pleasant. This paper demonstrates a system that provides pleasant route recommendation. Currently, we focus on routes that have much green and bright views. The system measures pleasure scores by extracting colors or objects in Google Street View panorama images and re-ranks shortest paths in the order of the computed pleasure scores. The current prototype provides route recommendation for city areas in Tokyo, Kyoto and San Francisco.