一种多模块图注意力机制和词库构建的电力敏感数据识别算法OACSTPCD

In the era of big data,there is a great risk of data sharing and fusion in the power industry. For this reason,this paper proposes a new text categorization algorithm based on multi-module graph attention mechanism. Firstly,the graph attention mechanism is combined with Transformer to introduce syntactic relations and use the graph attention mechanism to extract text features. Secondly,in order to increase the accuracy of sensitive word recognition and improve the recognition of sensitive words by the word separation tool,this paper proposes a new method for constructing sensitive word base. Experiments prove that the proposed method performs well in the sensitive text classification task and has good scalability at the same time.