A Knowledge-Enhanced Recommendation Model with Attribute-Level Co-Attention

Abstract

Deep neural networks (DNNs) have been widely employed in recommender systems including incorporating attention mechanism for performance improvement. However, most of existing attentionbased models only apply item-level attention on user side, restricting the further enhancement of recommendation performance. In this paper, we propose a knowledge-enhanced recommendation model ACAM, which incorporates item attributes distilled from knowledge graphs (KGs) as side information, and is built with a co-attention mechanism on attribute-level to achieve performance gains. Specifically, each user and item in ACAM are represented by a set of attribute embeddings at first. Then, user representations and item representations are augmented simultaneously through capturing the correlations between different attributes by a co-attention module. Our extensive experiments over two realistic datasets show that the user representations and item representations augmented by attribute-level co-attention gain ACAM's superiority over the state-of-the-art deep models.

Evaluating Economy Hotel Website Service Quality: A Hybrid Bounded Rationality Behavioral Decision Support Model

“…In Figure 3, the x-axes of three sub-figures indicate the values of parameter t which was searched from 1 to 10 under three LSFs ( f ∈{1, 2, 3}), respectively. In Figure 4, two parameters α and r in the new distance measure are investigated under [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1] and [1][2][3][4][5][6][7][8][9][10][11], respectively, where the LSF is fixed as f = 1. Overall, the sensitivity analysis results confirm that our proposed method is competent and stable when evaluating the service quality of economy hotel websites.…”

“…The multi-experts MCDM processes generally involve two main phases: criteria weight determination and alternative ranking generation. As for the first phase, existing MCDM methods often assume that the criteria are independent, which is not always true in real-world problems [5]. In the field of hotel website service quality evaluation, there are complex interrelationships among diverse evaluation dimensions and criteria.…”

Dual attention recommendation algorithm based on item attributes

“…ACAM [27] learned associations between different attributes through a coattention module as a way to improve recommendation performance. AANMF [28] employed the attention mechanism in learning the weights between different attributes,and obtained the user-item matching degree through the neural matrix decomposition network.…”

Exploiting side information for effective recommendation

“…For instance, KTUP [13] jointly learns the recommendation model and the KG completion task based on TransH [137]. ACAM [154] incorporates the KG embedding task via TransH to learn better item attribute embeddings. RCF [150] models the user preferences and item relations via DistMult [152].…”

“…Model direct relations between entities for embedding learning Fail to capture high-order user-item connectivities KTUP [13], ACAM [154], RCF [150], JNSKR [17], KERL [127], MKM-SR [79], Chorus [124] Path-based Methods Explore linear paths between a user-item pair to represent their connectivity Fail to exploit both semantics and topology of KGs PER [159], SimMF [101], HERec [102], HINE [47], HAN [135], HCDIR [7], ACKRec [32], MetaHIN [66], RKGE [111], KPRN [132], KARN [173], ADAC [165],…”