Structuring ontology terms automatically based on lexical granularity and machine learning (Preprint)

BACKGROUND As the manual creation and maintenance of biomedical ontologies are labor-intensive, automatic aids are desirable in the life cycle of ontology development. OBJECTIVE In this study, provided with a set of concept names in the Foundational Model of Anatomy (FMA), we aim to propose an innovative method for automatically generating the taxonomy and the partonomy structures among them, respectively. METHODS Our approach comprises two main tasks: The first task is predicting the direct relation between two given concept names, by utilizing word embedding and training machine learning models Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory Networks (Bi-LSTM). The second task is introducing an original granularity-based method to identify the semantic structures among a group of given concept names, by leveraging the trained models above. RESULTS Results show that both CNN and Bi-LSTM perform well on the first task, with F1 measures above 0.91. For the second task, our approach achieves an average F1 measure of 0.79 on 100 case studies in FMA using Bi-LSTM, which outperforms the primitive pairwise-based method. CONCLUSIONS In conclusion, we investigate an automatic way to predict a hierarchical relation between two concept names, based on which, we further invent a methodology to structure a group of concept names automatically. This study is an initial investigation that will shed light on further work about automatic creation and enrichment of biomedical ontologies. CLINICALTRIAL