基于双通道特征融合的WPOS-GRU专利分类方法

为提高专利文本自动分类的效率和准确度，提出一种基于双通道特征融合的WPOS-GRU（word2vec and part of speech gated recurrent unit）专利文本自动分类方法。首先获取专利摘要文本，并进行清洗和预处理；然后对专利文本进行词向量表示和词性标注，并将专利文本分别映射为word2vec词向量序列和POS词性序列；最后使用两种特征通道训练WPOS-GRU模型，并对模型效果进行实验分析。通过对比传统专利分类方法和单通道专利分类方法，双通道特征融合的WPOS-GRU专利分类方法提高了分类效果。提出的方法节省了大量的人力成本，提高了专利文本分类的准确度，更能满足大量专利文本分类任务自动化高效率的需要。

WPOS-GRU patent classification method based on two-channel feature fusion

In order to improve the efficiency and accuracy of patent text automatic classification, this paper proposed a WPOS-GRU patent text automatic classification method based on two-channel feature fusion. Firstly, this method obtained, cleaned and pretreated the patent summary text, then represented the patent text by word vector and part-of-speech tagging, and mapped the patent text into word2vec word vector sequence and POS part-of-speech sequence respectively. Finally, this paper trained WPOS-GRU model by two feature channels, and analyzed experimentally the effect of the model. By comparing the traditional patent classification method with the single-channel patent classification method, the WPOS-GRU patent classification method based on two-channel feature fusion improves the classification effect. The proposed method saves a lot of manpower costs, improves the accuracy of patent text classification, and can meet the needs of automation and high efficiency of a large number of patent text classification tasks.