Question Answering Based Assisted Decision for Electric Power Fault Diagnosis

With the development of Information Technology, question answering approaches have been applied in assisted decision, which can assist custom service staffs to automatically answer some basic questions and therefore can greatly reduce the cost of system operation. In this paper, we propose a question answering based approach for electric power fault diagnosis. First, the question answer data is analyzed and preprocessed from the conversation of manually answering the possible causes of given power equipment faults for making sure that one question corresponds to multiple answers. Then, the question and answers at word-level are matched based on a general “compare-aggregate” framework. Based on the matching measure, we can further rank the answers and select the answer with the highest correlation as the most likely fault cause for assisted decision. The relevant experimental results show that the model achieves good results and finds the best answer to the question.