【摘要】 文章提出了一种基于主题分类与语义相似度的专利推荐算法。该算法通过引入神经网络Bert,将专利标题及摘要进行关键词提取与词向量转换,使用DBSCAN聚类方法根据词向量构建专利主题领域类别,与文本相似度框架SimNet结合构成一个整体分析模型,进而将待预测专利文本输入已训练好的分析模型中,以此进行专利推荐。文章提出的算法在专利推荐方面可以获得更好的推荐效果。更多还原

【Abstract】 This paper proposes a patent recommendation algorithm based on topic classification and semantic similarity. By using the neural network Bert algorithm, this algorithm extracted keywords and transformed word vectors from patent titles and abstracts, and then DBSCAN clustering method was used to construct patent topic domain categories according to word vectors, which were combined with text similarity framework SimNet to constitute a whole analysis model, finally, the patent text to be predicted is input into the trained analysis model to make patent recommendation. The algorithm proposed in this paper can achieve better recommendation effect in patent recommendation.更多还原