In the existing recommender systems, matrix factorization (MF) is widely applied to model user preferences and item features by mapping the user-item ratings into a low-dimension latent vector space. However, MF has ignored the individual diversity where the user's preference for different unrated items is usually different. A fixed representation of user preference factor extracted by MF cannot model the individual diversity well, which leads to a repeated and inaccurate recommendation. To this end, we propose a novel latent factor model called adaptive deep latent factor model (ADLFM), which learns the preference factor of users adaptively in accordance with the specific items under consideration. We propose a novel user representation method that is derived from their rated item descriptions instead of original user-item ratings. Based on this, we further propose a deep neural networks framework with an attention factor to learn the adaptive representations of users. Extensive experiments on Amazon data sets demonstrate that ADLFM outperforms the state-of-the-art baselines greatly. Also, further experiments show that the attention factor indeed makes a great contribution to our method.

Measuring word semantic similarity is a generic problem with a broad range of applications such as ontology mapping, computational linguistics and artificial intelligence. Previous approaches to computing word semantic similarity did not consider concept occurrence frequency and word's sense number. This paper introduced Hyponymy graph, and based on which proposed a novel word semantic similarity model. For two words to be compared, we first retrieve their related concepts; then produce lowest common ancestor matrix and distance matrix between concepts; finally calculate distance-based similarity and information-based similarity, which are integrated to get final semantic similarity. The main contribution of our method is that both concept occurrence frequency and word's sense number are taken into account. This similarity measurement more closely fits with human rating and effectively simulates human thinking process. Our experimental results on benchmark dataset M&C and R&G with WordNet2.1 as platform demonstrate roughly 0.9%–1.2% improvements over existing best approaches.