Word embedding and deep learning methods have shown to be very effective
in automated text information mining. In this work, a chemical accident
case analysis method is proposed and tested based on the methods.
Firstly, word vectors for the text corpus of chemical accident cases
were produced based on word2vec. And then Bidirectional LSTM model with
attention mechanism for text classification was constructed. Finally,
case studies on the correlation analysis of the common trends of
chemical accidents and automated text classification of chemical
accident cases documents were performed respectively. The results
revealed that the proposed method can identify common principles of
chemical accidents and classify chemical accident cases. These findings
highlight the feasibility of chemical accident case information mining
based on word embedding through case studies. Compared with rule-based
information mining, our method improves the efficiency, automation, and
intelligence of information mining from chemical accident cases
documents.