Hierarchical topic model has been widely applied in many real applications, because it can build a hierarchy on topics with guaranteeing of topics' quality. Most of traditional methods build a hierarchy by adopting l...Hierarchical topic model has been widely applied in many real applications, because it can build a hierarchy on topics with guaranteeing of topics' quality. Most of traditional methods build a hierarchy by adopting low-level topics as new features to construct high-level ones, which will often cause semantic confusion between low-level topics and high-level ones. To address the above problem, we propose a novel topic model named hierarchical sparse NMF with orthogonal constraint (HSOC), which is based on non-negative matrix factorization and builds topic hierarchy via splitting super-topics into sub-topics. In HSOC, we introduce global independence, local independence and information consistency to constraint the split topics. Extensive experimental results on real-world corpora show that the purposed model achieves comparable performance on topic quality and better performance on semantic feature representation of documents compared with baseline methods.展开更多
User-generated content(UGC) such as blogs and twitters are exploding in modern Internet services. In such systems, recommender systems are needed to help people filter vast amount of UGC generated by other users. Howe...User-generated content(UGC) such as blogs and twitters are exploding in modern Internet services. In such systems, recommender systems are needed to help people filter vast amount of UGC generated by other users. However, traditional recommendation models do not use user authorship of items. In this paper, we show that with this additional information, we can significantly improve the performance of recommendations. A generative model that combines hierarchical topic modeling and matrix factorization is proposed. Empirical results show that our model outperforms other state-of-the-art models, and can provide interpretable topic structures for users and items. Furthermore, since user interests can be inferred from their productions, recommendations can be made for users that do not have any ratings to solve the cold-start problem.展开更多
文摘Hierarchical topic model has been widely applied in many real applications, because it can build a hierarchy on topics with guaranteeing of topics' quality. Most of traditional methods build a hierarchy by adopting low-level topics as new features to construct high-level ones, which will often cause semantic confusion between low-level topics and high-level ones. To address the above problem, we propose a novel topic model named hierarchical sparse NMF with orthogonal constraint (HSOC), which is based on non-negative matrix factorization and builds topic hierarchy via splitting super-topics into sub-topics. In HSOC, we introduce global independence, local independence and information consistency to constraint the split topics. Extensive experimental results on real-world corpora show that the purposed model achieves comparable performance on topic quality and better performance on semantic feature representation of documents compared with baseline methods.
基金Project supported by the Monitoring Statistics Project on Agricultural and Rural Resources,MOA,Chinathe Innovative Talents Project,MOA,Chinathe Science and Technology Innovation Project Fund of Chinese Academy of Agricultural Sciences(No.CAAS-ASTIP-2015-AI I-02)
文摘User-generated content(UGC) such as blogs and twitters are exploding in modern Internet services. In such systems, recommender systems are needed to help people filter vast amount of UGC generated by other users. However, traditional recommendation models do not use user authorship of items. In this paper, we show that with this additional information, we can significantly improve the performance of recommendations. A generative model that combines hierarchical topic modeling and matrix factorization is proposed. Empirical results show that our model outperforms other state-of-the-art models, and can provide interpretable topic structures for users and items. Furthermore, since user interests can be inferred from their productions, recommendations can be made for users that do not have any ratings to solve the cold-start problem.