1.Cross-domain Recommendation Without Sharing User-relevant Data

研究方向：cross-domain recommendation Task

Goal: combine data from different websites to improve recommendation task

Challenge: Despite many research efforts on this task, the main drawback is that they largely assume the data of different systems can be fully shared.

methods: NATR(short for Neural Attentive Transfer Recommendation) To avoid the leak of user privacy during the data sharing process, it consider sharing only the information of the item side, rather than user behavior data. Specifically, we transfer the item embeddings across domains, making it easier for two companies to reach a consensus (e.g., legal policy) on data sharing since the data to be shared is user-irrelevant and has no explicit semantics.

step:

Our proposed solution, which has three steps, is illustrated in Figure 1.

In the first step, an embedding-based recommender model, MF for example, is trained on the user-item interaction matrix of the auxiliary domain to obtain item embeddings.
In the second step, item embeddings of the auxiliary domain are sent to the target domain; note that only the embeddings of overlapped items are necessary to be sent, which are subjected to the data-sharing policy between two companies.
Finally, the target domain trains a recommender model with the consideration of the transferred item embeddings.

contribution

We present a new paradigm for cross-domain recommendation without sharing user-relevant data, in which only item-side data can be shared across domains. To allow the transferring of CF signal, we propose to share the item embeddings which are learned from user-item interactions of the auxiliary domain.
We propose a new solution NATR to resolve the key challenges in leveraging transferred item embeddings. The twolevel attention design allows NATR to distill useful signal from transferred item embeddings, and appropriately combine them with the data of the target domain.
We conduct extensive experiments on two real-world datasets to demonstrate our proposed method. More ablation studies verify the efficacy of our designed components, and the utility of transferred item embeddings in addressing the data sparsity issue.

framework

2.Dual Graph Attention Networks for Deep Latent Representation of Multifaceted Social Effects in Recommender Systems

Abstract

contribution

framework

3.Exploiting Ratings, Reviews and Relationships for Item Recommendations in Topic Based Social Networks

Abstract

Many e-commerce platforms today allow users to give their rating scores and reviews on items as well as to establish social relationships with other users. As a result, such platforms accumulate heterogeneous data including numeric scores, short textual reviews, and social relationships. HHowever, many recommender systems only consider historical user feedbacks in modeling user preferences. More specifically, most existing recommendation approaches only use rating scores but ignore reviews and social relationships in the user-generated data. In this paper, we propose TSNPF—a latent factor model to effectively capture user preferences and item features. Employing Poisson factorization, TSNPF fully exploits the wealth of information in rating scores, review text and social relationships altogether. It extracts topics of items and users from the review text and makes use of similarities between user pairs with social relationships, which results in a comprehensive understanding of user preferences. Experimental results on real-world datasets demonstrate that our TSNPF approach is highly effective at recommending items to users.

contribution

We propose a method based on Gamma-Poisson distribution to extract the topic intensities of items and users from usergenerated textual reviews. Compared to previous techniques, our method is able to address the usual problem of data scarcity.
We propose TSNPF, a conjugate graphical model based on Poisson factorization which only models non-zero observations in ratings, reviews and social relations simultaneously via interpretable user preferences and item attributes. In addition, we propose a closed form mean-field variational inference method to train TSNPF.
We evaluate the performance of TSNPF using three publicly available real datasets. The results show that TSNPF outperforms state-of-the-art alternatives.

framework

主题提取与推荐系统的结合

4.GhostLink: Latent Network Inference for Influence-aware Recommendation.(introduction 写的不错)

Abstract (probabilistic graphical model)

Social influence plays a vital role in shaping a user’s behavior in online communities dealing with items of fine taste like movies, food, and beer. For online recommendation, this implies that users’ preferences and ratings are influenced due to other individuals. Given only time-stamped reviews of users, can we find out whoinfluences- whom, and characteristics of the underlying influence network? Can we use this network to improve recommendation?

While prior works in social-aware recommendation have leveraged social interaction by considering the observed social network of users, many communities like Amazon, Beeradvocate, and Ratebeer do not have explicit user-user links.Therefore,we propose GhostLink, an unsupervised probabilistic graphical model, to automatically learn the latent influence network underlying a review community – given only the temporal traces (timestamps) of users’ posts and their content. Based on extensive experiments with four real-world datasets with 13 million reviews, we show that GhostLink improves item recommendation by around 23% over state-of-the-art methods that do not consider this influence. As additional use-cases, we show that GhostLink can be used to differentiate between users’ latent preferences and influenced ones, as well as to detect influential users based on the learned influence graph.

contribution

We propose an unsupervised probabilistic generative model GhostLink based on Latent Dirichlet Allocation to learn a latent influence graph in online communities without requiring explicit user-user links or a social network. This is the first work that solely relies on timestamped review data.
We propose an efficient algorithm based on Gibbs sampling to estimate the hidden parameters in GhostLink that empirically demonstrates fast convergence.
We perform large-scale experiments in four communities with 13 million reviews, 0.5 mil. items, and 1 mil. users where we show improved recommendation for item rating prediction by around 23% over state-of-the-art methods. Moreover, we analyze the properties of the influence graph and use it for use-cases like finding influential members in the community.

framework

5.Graph Neural Networks for Social Recommendation

Abstract

In recent years, Graph Neural Networks (GNNs), which can naturally integrate node information and topological structure, have been demonstrated to be powerful in learning on graph data. These advantages of GNNs provide great potential to advance social recommendation since data in social recommender systems can be represented as user-user social graph and user-item graph; and learning latent factors of users and items is the key.

However, building social recommender systems based on GNNs faces challenges. For example, the user-item graph encodes both interactions and their associated opinions; social relations have heterogeneous strengths; users involve in two graphs (e.g., the useruser social graph and the user-item graph).

To address the three aforementioned challenges simultaneously, in this paper,we present a novel graph neural network framework (GraphRec) for social recommendations. In particular, we provide a principled approach to jointly capture interactions and opinions in the user-item graph and propose the framework GraphRec, which coherently models two graphs and heterogeneous strengths. Extensive experiments on two real-world datasets demonstrate the effectiveness of the proposed framework GraphRec.

contribution

We propose a novel graph neural network GraphRec, which can model graph data in social recommendations coherently;
We provide a principled approach to jointly capture interactions and opinions in the user-item graph;
We introduce a method to consider heterogeneous strengths of social relations mathematically;
We demonstrate the effectiveness of the proposed framework on various real-world datasets.

framework

performance

6.Hierarchical Temporal Convolutional Networks for Dynamic Recommender Systems(工程应用)

Abstract

Recommender systems that can learn from cross-session data to dynamically predict the next item a user will choose are crucial for online platforms. However, existing approaches often use out-ofthe-box sequence models which are limited by speed and memory consumption, are often infeasible for production environments, and usually do not incorporate cross-session information, which is crucial for effective recommendations.

Here we propose Hierarchical Temporal Convolutional Networks (HierTCN), a hierarchical deep learning architecture that makes dynamic recommendations based on users’ sequential multi-session interactions with items. HierTCN is designed for web-scale systems with billions of items and hundreds of millions of users. It consists of two levels of models: The high-level model uses Recurrent Neural Networks (RNN) to aggregate users’ evolving long-term interests across different sessions, while the low-level model is implemented with Temporal Convolutional Networks (TCN), utilizing both the long-term interests and the short-term interactions within sessions to predic the next interaction.

contribution

HierTCN has a significant performance improvement over existing deep learning models by about 30% on a public XING dataset and 18% on a private large-scale Pinterest dataset.
Compared with RNN-based approaches, HierTCN is 2.5 times faster in terms of training time and allows for much easier gradient backpropagation.
Compared with CNN-based approaches, HierTCN requires roughly 10% data memory usage and allows for easy latent feature extraction.

framework

7.How Intention Informed Recommendations Modulate Choices: A Field Study of SpokenWord Content

Abstract

People’s content choices are ideally driven by their intentions, aspirations, and plans.

However, in reality, choices may be modulated by recommendation systems which are typically trained to promote popular items and to reinforce users’ historical behavior. As a result, the utility and user experience of content consumption can be affected implicitly and undesirably.

To study this problem, we conducted a 2 × 2 randomized controlled field experiment (105 urban college students) to compare the effects of intention informed recommendations with classical intention agnostic systems. The study was conducted in the context of spokenwordweb content (podcasts) which is often consumed through subscription sites or apps. We modified a commercial podcast app to include (1) a recommender that takes into account users’ stated intentions at onboarding, and (2) a Collaborative Filtering (CF) recommender during daily use. Our study suggests that: (1) intention-aware recommendations can significantly raise users’ interactions (subscriptions and listening) with channels and episodes related to intended topics by over 24%, even if such a recommender is only used during onboarding, and (2) the CF-based recommender doubles users’ explorations on episodes from not-subscribed channels and improves satisfaction for users onboarded with the intention-aware recommender.

contribution

framework

8.How Serendipity Improves User Satisfaction with Recommendations? A Large-Scale User Evaluation

Abstract

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.

9.Improving Outfit Recommendation with Co-supervision of Fashion Generation (图像：衣服类)

Abstract

The task of fashion recommendation includes twomain challenges: visual understanding and visual matching. Visual understanding aims to extract effective visual features. Visual matching aims to model a human notion of compatibility to compute a match between fashion items. Most previous studies rely on recommendation loss alone to guide visual understanding and matching. Although the features captured by thesemethods describe basic characteristics (e.g., color, texture, shape) of the input items, they are not directly related to the visual signals of the output items (to be recommended). This is problematic because the aesthetic characteristics (e.g., style, design), based on which we can directly infer the output items, are lacking. Features are learned under the recommendation loss alone, where the supervision signal is simply whether the given two items are matched or not.

To address this problem, we propose a neural co-supervision learning framework, called the FAshion RecommendationMachine (FARM). FARM improves visual understanding by incorporating the supervision of generation loss, which we hypothesize to be able to better encode aesthetic information. FARMenhances visual matching by introducing a novel layer-to-layer matching mechanism to fuse aesthetic information more effectively, and meanwhile avoiding paying too much attention to the generation quality and ignoring the recommendation performance.

Extensive experiments on two publicly available datasets show that FARM outperforms state-of-the-art models on outfit recommendation, in terms of AUC and MRR. Detailed analyses of generated and recommended items demonstrate that FARM can encode better features and generate high quality images as references to improve recommendation performance.

contribution

We propose a neural co-supervision learning framework, FARM, for outfit recommendation that simultaneously yields recommendation and generation.
We propose a layer-to-layer matching mechanism that acts as a bridge between generation and recommendation, and improves recommendation by leveraging generation features.
Our proposed approach is shown to be effective in experiments on two large-scale datasets.

framework

10.Jointly Learning Explainable Rules for Recommendation with Knowledge Graph

Abstract

Explainability and effectiveness are two key aspects for building recommender systems. Prior efforts mostly focus on incorporating side information to achieve better recommendation performance.

However, these methods have some weaknesses: (1) prediction of neural network-based embedding methods are hard to explain and debug; (2) symbolic, graph-based approaches (e.g., meta path-based models) require manual efforts and domain knowledge to define patterns and rules, and ignore the item association types (e.g. substitutable and complementary).

In this paper, we propose a novel joint learning framework to integrate induction of explainable rules from knowledge graph with construction of a rule-guided neural recommendation model. The framework encourages two modules to complement each other in generating effective and explainable recommendation:

inductive rules, mined from item-centric knowledge graphs, summarize common multi-hop relational patterns for inferring different item associations and provide human-readable explanation for model prediction; 2) recommendation module can be augmented by induced rules and thus have better generalization ability dealing with the cold-start issue.

Extensive experiments1 show that our proposed method has achieved significant improvements in item recommendation over baselines on real-world datasets. Our model demonstrates robust performance over “noisy" item knowledge graphs, generated by linking item names to related entities.

contribution

We utilize a large-scale knowledge graph to derive rules between items from item associations.
We propose a joint optimization framework that induces rules from knowledge graphs and recommends items based on the rules at the same time.
We conduct extensive experiments on real-world datasets. Experimental results prove the effectiveness of our framework in accurate and explainable recommendation.

framework

11.Jointly Leveraging Intent and Interaction Signals to Predict User Satisfaction with Slate Recommendations.

Abstract

Detecting and understanding implicit measures of user satisfaction are essential for enhancing recommendation quality. When users interact with a recommendation system, they leave behind fine grained traces of interaction signals, which contain valuable information that could help gauging user satisfaction. User interaction with such systems is often motivated by a specific need or intent, often not explicitly specified by the user, but can nevertheless inform on how the user interacts with, and the extent to which the user is satisfied by the recommendations served. In this work, we consider a complex recommendation scenario, called Slate Recommendation, wherein a user is presented with an ordered set of collections, called slates, in a specific page layout. We focus on the context of music streaming and leverage fine-grained user interaction signals to tackle the problem of predicting user satisfaction.

We hypothesize that user interactions are conditional on the specific intent users have when interacting with a recommendation system, and highlight the need for explicitly considering user intent when interpreting interaction signals. We present diverse approaches to identify user intents (interviews, surveys and a quantitative approach) and identify a set of common intents users have in a music streaming recommendation setting. Additionally, we identify the importance of shared learning across intents and propose a multi-level hierarchical model for user satisfaction prediction that leverages user intent information alongside interaction signals. Our findings from extensive experiments on a large scale real world data demonstrate (i) the utility of considering different interaction signals, (ii) the role of intents in interpreting user interactions and (iii) the interplay between interaction signals and intents in predicting user satisfaction.

12.Modeling Heart Rate and Activity Data for Personalized Fitness Recommendation(健康推荐)

Abstract

Activity logs collected from wearable devices (e.g. Apple Watch, Fitbit, etc.) are a promising source of data to facilitate a wide range of applications such as personalized exercise scheduling, workout recommendation, and heart rate anomaly detection.

However, such data are heterogeneous, noisy, diverse in scale and resolution, and have complex interdependencies, making them challenging to model.

In this paper, we develop context-aware sequential models to capture the personalized and temporal patterns of fitness data.

Specifically, we propose FitRec – an LSTM-based model that captures two levels of context information: context within a specific activity, and context across a user’s activity history.

contribution

framework

13.Modeling Item-Specific Temporal Dynamics of Repeat Consumption for Recommender Systems

Abstract

Repeat consumption is a common scenario in daily life, such as repurchasing items and revisiting websites, and is a critical factor to be taken into consideration for recommender systems. Temporal dynamics play important roles in modeling repeat consumption. It is noteworthy that for items with distinct lifetimes, consuming tendency for the next one fluctuates differently with time. For example, users may repurchase milk weekly, but it is possible to repurchase mobile phone after a long period of time. Therefore, how to adaptively incorporate various temporal patterns of repeat consumption into a holistic recommendation model has been a new and important problem.

In this paper, we propose a novel unified model with introducing Hawkes Process into Collaborative Filtering (CF). Different from most previous work which ignores various time-varying patterns of repeat consumption, the model explicitly addresses two item-specific temporal dynamics: (1) short-term effect and (2) lifetime effect, which is named as Short-Term and Life-Time Repeat Consumption (SLRC) model. SLRC learns importance of the two factors for each item dynamically by interpretable parameters.

contribution

14.Multi-Task Feature Learning for Knowledge Graph Enhanced Recommendation(cross-domain recommendation)

Abstract

Collaborative filtering often suffers from sparsity and cold start problems in real recommendation scenarios, therefore, researchers and engineers usually use side information to address the issues and improve the performance of recommender systems.

In this paper, we consider knowledge graphs as the source of side information.

We propose MKR, a Multi-task feature learning approach for Knowledge graph enhanced Recommendation. MKR is a deep end-to-end framework that utilizes knowledge graph embedding task to assist recommendation task.

contribution

framework

15.Multimodal Review Generation for Recommender Systems

Abstract

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

contribution

We design the MRG model (see Section 3), which jointly models rating prediction and text generation at the review level by incorporating LSTM cells with a novel fusion gate as a kind of soft attention to weigh the relative contributions of sentiment features and visual features that provide context to the text generation. We also describe the learning and inference algorithms respectively.

framework

16.Personalized Bundle List Recommendation

Abstract

Product bundling, offering a combination of items to customers, is one of the marketing strategies commonly used in online ecommerce and offline retailers. A high-quality bundle generalizes frequent items of interest, and diversity across bundles boosts the user-experience and eventually increases transaction volume.

In this paper, we formalize the personalized bundle list recommendation as a structured prediction problem and propose a bundle generation network (BGN), which decomposes the problem into quality/diversity parts by the determinantal point processes (DPPs). BGN uses a typical encoder-decoder framework with a proposed feature-aware softmax to alleviate the inadequate representation of traditional softmax, and integrates the masked beam search and DPP selection to produce high-quality and diversified bundle list with an appropriate bundle size.