Joint Representation Learning for Top-N Recommendation with Heterogeneous Information Sources

Zhang, Yongfeng; Ai, Qingyao; Chen, Xu; Croft, W. Bruce

doi:10.1145/3132847.3132892

Cited by 275 publications

(215 citation statements)

References 36 publications

Supporting

Mentioning

199

Contrasting

Order By: Relevance

“…Finding an explanation path, however, is difficult for an arbitrary (u, i) pair. Because we only observe a limited number of relationship triples in the training data, the knowledge graph built on product data usually is sparse [6,69,71] . In most cases, it is impossible to find two sets of relationships {r k u } and {r j i } that directly link the user u to the item i.…”

Section: Extraction Algorithmmentioning

confidence: 99%

Explainable Product Search with a Dynamic Relation Embedding Model

Zhang

et al. 2019

ACM Trans. Inf. Syst.

Self Cite

View full text Add to dashboard Cite

USA Product search is one of the most popular methods for customers to discover products online. Most existing studies on product search focus on developing effective retrieval models that rank items by their likelihood to be purchased. They, however, ignore the problem that there is a gap between how systems and customers perceive the relevance of items. Without explanations, users may not understand why product search engines retrieve certain items for them, which consequentially leads to imperfect user experience and suboptimal system performance in practice. In this work, we tackle this problem by constructing explainable retrieval models for product search. Specifically, we propose to model the "search and purchase" behavior as a dynamic relation between users and items, and create a dynamic knowledge graph based on both the multi-relational product data and the context of the search session. Ranking is conducted based on the relationship between users and items in the latent space, and explanations are generated with logic inferences and entity soft matching on the knowledge graph. Empirical experiments show that our model, which we refer to as the Dynamic Relation Embedding Model (DREM), significantly outperforms the state-of-the-art baselines and has the ability to produce reasonable explanations for search results.

show abstract

Section: Extraction Algorithmmentioning

confidence: 99%

Explainable Product Search with a Dynamic Relation Embedding Model

Zhang

et al. 2019

ACM Trans. Inf. Syst.

Self Cite

View full text Add to dashboard Cite

show abstract

“…As the goal of recommender system is to provide the target user with a ranking list of top few items, which are most likely to be preferred by the user, recommendation is rather a ranking problem instead of rating. In light of this, MF moves to model the relative preferences between different items and the pairwise learning approach has been widely adopted to achieve the goal [35,54]. In pairwise learning, the user and item vectors are learnt by settingr u,i >r u,k for any two pairs that satisfy (u, i) ∈ R and (u, k) R. A typical example is the bayesian personalized ranking (BPR) [35].…”

Section: Our Proposed Model 31 Notations and Backgroundmentioning

confidence: 99%

“…For each user-item (u, i) pair, our model computes a weight vector a u,i ∈ R f to indicate the importance of i's aspects for u. In addition, the side information of items is exploited to estimate the weight vector, as side information conveys rich features of items, especially text reviews and item images, which are wellrecognized to provide notable and complementary features of items in different aspects [9,54]. We adopt the recent advancement of attention mechanism [6,10] to estimate the attention vector.…”

Section: Multimodal Attentive Metric Learningmentioning

confidence: 99%

“…Different types of side information are considered to be complementary to each other. With the advancement of representation learning [3,45], several methods have been proposed to exploit multi-source information for recommendation, such as the Collaborative Knowledge based Embedding (CKE) [51] method using structure information, textual and visual knowledge, as well as the Joint Representation Learning (JRL) method [54] exploiting ratings, reviews and item images.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Proceedings of the 27th ACM International Conference on Multimedia

2019

View full text Add to dashboard Cite

Most existing recommender systems represent a user's preference with a feature vector, which is assumed to be fixed when predicting this user's preferences for different items. However, the same vector cannot accurately capture a user's varying preferences on all items, especially when considering the diverse characteristics of various items. To tackle this problem, in this paper, we propose a novel Multimodal Attentive Metric Learning (MAML) method to model user diverse preferences for various items. In particular, for each user-item pair, we propose an attention neural network, which exploits the item's multimodal features to estimate the user's special attention to different aspects of this item. The obtained attention is then integrated into a metric-based learning method to predict the user preference on this item. The advantage of metric learning is that it can naturally overcome the problem of dot product similarity, which is adopted by matrix factorization (MF) based recommendation models but does not satisfy the triangle inequality property. In addition, it is worth mentioning that the attention mechanism cannot only help model user's diverse preferences towards different items, but also overcome the geometrically restrictive problem caused by collaborative metric learning. Extensive experiments on large-scale real-world datasets show that our model can substantially outperform the state-of-the-art baselines, demonstrating the potential of modeling user diverse preference for recommendation.

show abstract

“…Focused learning technique [32] improves the recommendations in terms of prediction accuracy for the cold-start items based on the objective of the customized matrix factorization and hyper parameter optimization. A Joint Representation Learning (JRL) framework [33] enhances the top-N recommendation with several heterogeneous information sources such as review text, numerical rating, product image, and so on. It can learn a complex prediction network for fast online calculation in the recommendation model.…”

Section: Ranking Based Recommender Systemsmentioning

confidence: 99%

Diversity-Ensured Semantic Movie Recommendation by Applying Linked Open Data

Srinivasan¹,

Mani²

2018

IJIES

View full text Add to dashboard Cite

Abstract:The recommender system becomes a significant research area due to the popularity of the social web. Traditional semantic recommender systems deliver poor performance when balancing the recommendation accuracy and diversity. Also, the rank-based recommendation methods lack to obtain the coverage of the entire preferences of the user in the top-N recommendation list. Thus, this paper presents the Diversity-Ensured Semantic-aware Item REcommendation (DESIRE) that deals with the consistent and reliable knowledge source to significantly improve the quality and provide the diversity-ensured top-N recommendation list. The DESIRE approach builds the semantically relevant graphs such as movie-centric and user rating-centric graph with the help of both the Linked Open Data (LOD) and the explicit ratings of the users. By extracting the semantic-path based features from the user rating-centric graph, it executes the ranking algorithm for the top-N movie recommendation. Moreover, the diversity-aware re-ranking tends to maintain the trade-off between the diversity and accuracy in the top-N recommendation.

show abstract

Joint Representation Learning for Top-N Recommendation with Heterogeneous Information Sources

Cited by 275 publications

References 36 publications

Explainable Product Search with a Dynamic Relation Embedding Model

Explainable Product Search with a Dynamic Relation Embedding Model

Proceedings of the 27th ACM International Conference on Multimedia

Diversity-Ensured Semantic Movie Recommendation by Applying Linked Open Data

Contact Info

Product

Resources

About