Research on Label Recognition Method of Power Grid Item Based on Deep Learning

This paper proposes an innovative method of item label construction based on knowledge modeling. The model is based on BiLSTM-CRF sequence labeling model to identify the label of short power grid item. At the same time, Word2vec+SVM model and multi-label classification technology are used to extract and model long power grid item labels. Improve the knowledge extraction for the system mining model of company item. At the same time, on the basis of a large number of item structured data, this paper uses supervised algorithm to automatically mine the item system based on a small amount of labeled data. The purpose of this paper is to manage the existing item by multi-dimensional and multi-level tag system, and to dynamically adjust and configure the trained model and support the dynamic expansion of rules. This method can adapt the data of new item well.