Semi-supervised Collaborative Filtering by Text-enhanced Domain Adaptation

Abstract

Data sparsity is an inherent challenge in the recommender systems, where most of the data is collected from the implicit feedbacks of users. This causes two difficulties in designing effective algorithms: first, the majority of users only have a few interactions with the system and there is no enough data for learning; second, there are no negative samples in the implicit feedbacks and it is a common practice to perform negative sampling to generate negative samples. However, this leads to a consequence that many potential positive samples are mislabeled as negative ones and data sparsity would exacerbate the mislabeling problem. To solve these difficulties, we regard the problem of recommendation on sparse implicit feedbacks as a semi-supervised learning task, and explore domain adaption to solve it. We transfer the knowledge learned from dense data to sparse data and we focus on the most challenging case âĂŤ there is no user or item overlap. In this extreme case, aligning embeddings of two datasets directly is rather sub-optimal since the two latent spaces encode very different information. As such, we adopt domain-invariant textual features as the anchor points to align the latent spaces. To align the embeddings, we extract the textual features for each user and item and feed them into a domain classifier with the embeddings of users and items. The embeddings are trained to puzzle the classifier and textual features are fixed as anchor points. By domain adaptation, the distribution pattern in the source domain is transferred to the target domain. As the target part can be supervised by domain adaptation, we abandon negative sampling in target dataset to avoid label noise.

Sampler Design for Implicit Feedback Data by Noisy-label Robust Learning

“…Hwang et al [24] utilized both implicit and explicit feedback data to improve the quality of negative sampling. Yu et al [42] replaced negative sampling by transfer learning.…”

Multi-Sparse-Domain Collaborative Recommendation via Enhanced Comprehensive Aspect Preference Learning

“…Early methods for single-target CDR jointly factorize the rating matrices in relevant domains to generate the representations that can capture the common preferences of shared users Metric NDCG@5 NDCG@10 HR@5 HR@10 [8, 10,12,19]. Recently, many DNN based methods have been proposed for better preference capturing across domains [1, 3-5, 7, 9, 11, 16, 23, 24, 27], which often apply transfer learning techniques like domain adaptation [2,20,23] to transfer domaininvariant preferences from a source domain to a target domain.…”

“…Cross-domain recommendation (CDR) has been attracting increasing attention of researchers for its ability to alleviate the data sparsity problem in recommender systems [22,31]. The common idea of most existing CDR methods is to improve the recommendation performance on a sparse target domain by transferring the information of a relevant source domain with rich data [3,4,23,24,27], which is called single-target CDR [28]. In real world, however, almost every domain suffers from data sparsity problem due to its ubiquity, which may cause the existing single-target CDR methods to fail in finding a dense auxiliary domain to help the sparse target domain.…”

“…GAPE can generate gate control vectors to regulate the transfer of the user's complementary aspect preferences, by which MSDCR can enhance a user's domain-specific aspect preference in each domain. (3) To obtain a user's comprehensive preference in a domain, we also propose a Multi-Domain Adaptation Network (MDAN) for the domain-invariant aspect preference learning, which applies domain adaptation [23] to capture a user's preferences common to different domains. (4) Finally, we introduce a multi-task framework to MSDCR, by which MSDCR can be trained under the collaborative supervisions of multiple relevant domains and make multi-target recommendations for a user based on the user's learned comprehensive preference in each domain.…”

Inferring Multilingual Domain-Specific Word Embeddings From Large Document Corpora

“…The retrieval task is known since long ago [23] and has been largely addressed by the NLP community (e.g., [44]- [46]).…”