Named entity recognition in aircraft design field based on deep learning

Abstract

Aircraft design is a kind of knowledge-intensive work involving multi-disciplinary integration, which needs the support of a large amount of knowledge on aircraft design field (ADF). At the same time, a large number of technical documents about AD also accumulate rich aircraft design knowledge. If this knowledge can be extracted, it can be used to guide the intelligent design and maintenance of aircraft. In this paper, we conduct our research for the named entity recognition, which is an important step of knowledge graph construction in ADF. For the problem of knowledge dispersion in ADF and lacking of training dataset, we design a platform for data acquisition and processing, and corpus annotation by crowdsourcing. And a novel deep neural network model, named AR+BiLSTM+CRF, which combines attention mechanism, Ranger optimizer, bidirectional LSTM, and CRF, is proposed for named entity recognition in ADF. The experimental results show that AR+BiLSTM+CRF model has excellent performance for named entity recognition in ADF.