【摘要】 针对一些缺少参考对齐的本体匹配任务，提出一种基于深度无监督学习的匹配技术，通过对文本的上下文信息进行学习，提取到抽象文本特征，以此找到对齐。由于高维度输入会影响计算的效率，针对本体的多种描述构建CNN(convolutional neural network)模块并且和不同的RNN(recurrent neural network)串行连接实现特征降维，提出一种改进的基于BiLSTM(bidirectional long and short term memory neural network)的注意力机制来提取较好的抽象特征。提出一种多主导的对齐集成策略将本体不同层次的对齐进行合并，提高匹配的质量。实验在OAEI(ontology alignment evaluation initiative)的benchmark测试集上进行，提出方法的评价指标较高，并且和其它匹配系统作比较，高质量的对齐验证了所提方法具有一定的先进性和创新性。更多还原

【Abstract】 Aiming at some ontology matching tasks that lack reference alignment, a deep unsupervised learning based matching technique was proposed, the context information of the text was learned, and the abstract features of the text were extracted to find similar alignment. The high-dimensional input affected the computing efficiency, so different RNNs(recurrent neural network) were connected to CNN(convolutional neural network) in a serial way to achieve feature dimension reduction according to multiple descriptions of ontologies, and an improved BiLSTM(bidirectional long and short term memory neural network) based attention mechanism was proposed to extract better abstract features. A multi-dominant alignment aggregation strategy was proposed to integrate the alignments at different levels of ontology to improve the quality of matching. The experiment was carried out on the benchmark test set of the OAEI(ontology alignment evaluation initiative). The evaluation indexes are high. Compared with other matching systems, the high quality alignment verifies that the proposed method is advanced and innovative.更多还原