Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Item Silk Road: Recommending Items from Information Domains to Social Users (1706.03205v1)

Published 10 Jun 2017 in cs.IR, cs.AI, and cs.SI

Abstract: Online platforms can be divided into information-oriented and social-oriented domains. The former refers to forums or E-commerce sites that emphasize user-item interactions, like Trip.com and Amazon; whereas the latter refers to social networking services (SNSs) that have rich user-user connections, such as Facebook and Twitter. Despite their heterogeneity, these two domains can be bridged by a few overlapping users, dubbed as bridge users. In this work, we address the problem of cross-domain social recommendation, i.e., recommending relevant items of information domains to potential users of social networks. To our knowledge, this is a new problem that has rarely been studied before. Existing cross-domain recommender systems are unsuitable for this task since they have either focused on homogeneous information domains or assumed that users are fully overlapped. Towards this end, we present a novel Neural Social Collaborative Ranking (NSCR) approach, which seamlessly sews up the user-item interactions in information domains and user-user connections in SNSs. In the information domain part, the attributes of users and items are leveraged to strengthen the embedding learning of users and items. In the SNS part, the embeddings of bridge users are propagated to learn the embeddings of other non-bridge users. Extensive experiments on two real-world datasets demonstrate the effectiveness and rationality of our NSCR method.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Xiang Wang (279 papers)
  2. Xiangnan He (200 papers)
  3. Liqiang Nie (191 papers)
  4. Tat-Seng Chua (360 papers)
Citations (262)

Summary

An Analysis of "Item Silk Road: Recommending Items from Information Domains to Social Users"

The paper "Item Silk Road: Recommending Items from Information Domains to Social Users" proposes a novel approach to the cross-domain social recommendation problem, which involves recommending items from information domains to users within social networking services (SNSs). This task is differentiated from traditional cross-domain recommender systems by considering the heterogeneity between information-oriented platforms (like Trip.com and Amazon) and social-oriented platforms (such as Facebook and Twitter). The authors introduce the notion of bridge users, who have profiles on both types of platforms, as the critical link for facilitating these recommendations.

Methodology

The authors introduce the Neural Social Collaborative Ranking (NSCR) model, which integrates user-item interactions from information domains with user-user connections from social domains. The NSCR employs a neural network architecture to effectively map and connect the disparate data from these domains. The model includes two key components:

  1. Information Domain Modeling: The NSCR employs an attribute-aware recommender scheme based on Neural Collaborative Filtering (NCF). It enhances learning by incorporating both user and item attributes through a pairwise pooling operation, which captures interactions between a user, an item, and their respective attributes.
  2. Social Domain Modeling: The NSCR facilitates the propagation of user-item interactions across domains, leveraging bridge users as conduits for transmitting learned embeddings to non-bridge users through social connections. This is achieved by using a graph Laplacian to enforce smoothness across the social network, ensuring that connected social users have similar embeddings.

The NSCR’s use of deep learning allows for capturing complex interactions between users, items, and attributes. Its multi-layered structure enables the learning of higher-order interactions, thus enhancing the embeddings’ expressiveness.

Evaluation and Results

The NSCR model was empirically tested on two real-world data sets, involving data from Trip.com, Facebook, and Twitter. Results from these experiments demonstrated that NSCR significantly outperformed existing cross-domain recommendation models like SFM and SR with regards to both AUC and Recall@5 metrics. This outcome underscores the model's effectiveness in bridging the heterogeneous domains through the innovate use of deep learning techniques.

Implications and Future Directions

The paper opens up several avenues for further exploration. Practically, the NSCR model could be adapted to other types of domains where partial overlap through bridge entities can be identified, such as between different kinds of social media platforms or between distinct e-commerce platforms. Theoretically, the approach presents a framework that could potentially be enhanced further by integrating more sophisticated graph-based learning techniques or by expanding the attribute set used in modeling interactions.

For future developments, other methods of capturing user signals beyond bridge users—such as inferring potential connections through attribute similarities or leveraging machine learning techniques for cold start users—could be explored. Furthermore, since the current paper focuses on user-user and user-item interactions, incorporating more dynamic and less structured data sources, such as text or user-generated content, could enrich recommendation quality.

In summary, this paper offers a substantive contribution to the cross-domain recommendation field by presenting a robust model that effectively capitalizes on the limited but crucial bridge users to offer tailored item recommendations in disparate domains. Its empirical success is likely to stimulate additional efforts to address this complex but increasingly relevant space within recommendation research.