A Patent recommendation algorithm based on topic classification and semantic similarity
A Patent recommendation algorithm based on topic classification and semantic similarity
复制标题
一种基于主题分类和语义相似度的专利推荐算法
DOI：
10.1109/icwcsg53609.2021.00063
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Jiahe Zhang
中科院分区：
文献类型：
--
作者：
Xiaojuan Liu;Yunye Wan;XinBo Liu;Jiahe Zhang
专利推荐算法作为信息推送的重要手段，是解决当今海量数据信息过载的重要手段。然而传统的推荐算法存在无法充分利用用户信息、系统冷启动、数据矩阵稀疏等问题，因此本文提出一种基于主题分类和语义相似度的专利推荐算法。该算法通过引入Bert神经网络，从专利标题和摘要中提取关键词，然后将其转化为词向量。利用它们，算法采用DBSCAN聚类方法构建专利主题领域类别。与文本相似度框架SimNet结合，成为整体分析模型。将需要预测的专利文本输入训练好的分析模型中，即可进行专利推荐工作。与传统推荐算法相比，实验表明，本文提出的算法在专利推荐上能够获得更好的推荐效果。
Patent recommendation algorithms, as an important means of information push, are an important means of solving the information overload of today's massive data. However, traditional recommendation algorithms have problems such as the inability to make full use of user information, system cold start, and sparse data matrix, so this paper proposes a patented recommendation algorithm based on topic classification and semantic similarity. By introducing the Bert neural network, this algorithm extracts keywords from patent titles and abstracts, and then transforms them into word vectors. By using them, the algorithm uses the DBSCAN clustering method to construct patent subject area categories. Combining with SimNet, a text similarity framework, it becomes a holistic analysis model. Inputting patent text to be predicted into trained analysis model, then it can do patent recommendation work. Comparing with the traditional recommendation algorithm, the experiment shows that the algorithm proposed in this paper can obtain a better recommendation effect on the patent recommendation.