论文标题
部分可观测时空混沌系统的无模型预测
User recommendation system based on MIND dataset
论文作者
论文摘要
如今，这是研究人员和其他人实现自己的兴趣的一种非常重要的方法，因为它提供了满足其需求的简短解决方案。由于互联网上有很多信息，因此新闻推荐系统使我们能够过滤内容并按照他的欲望和兴趣成比例地将其交付给用户。 RSS具有三种技术：基于内容的过滤，协作过滤和混合过滤。我们将与系统一起使用Mind数据集，该系统于2019年收集，这是该数据集中的最大挑战，因为存在很多歧义和复杂的文本处理。在本文中，将介绍我们提出的推荐系统。我们系统的核心使用了手套算法来嵌入单词嵌入和表示。此外，多头注意力层计算单词的注意力，以生成推荐新闻列表。最后，我们比AUC 71.211，MRR 35.72，NDCG@5 38.05和NDCG@10 44.45中的其他相关作品获得了良好的成绩。
Nowadays, it's a very significant way for researchers and other individuals to achieve their interests because it provides short solutions to satisfy their demands. Because there are so many pieces of information on the internet, news recommendation systems allow us to filter content and deliver it to the user in proportion to his desires and interests. RSs have three techniques: content-based filtering, collaborative filtering, and hybrid filtering. We will use the MIND dataset with our system, which was collected in 2019, the big challenge in this dataset because there is a lot of ambiguity and complex text processing. In this paper, will present our proposed recommendation system. The core of our system we have used the GloVe algorithm for word embeddings and representation. Besides, the Multi-head Attention Layer calculates the attention of words, to generate a list of recommended news. Finally, we achieve good results more than some other related works in AUC 71.211, MRR 35.72, nDCG@5 38.05, and nDCG@10 44.45.
