Research Sharing-Oriented Functional Neuroimaging Named Entity Recognition

Named entity recognition is a basic and core task of information extraction on functional neuroimaging literatures. However, existing researches only focus on cognitive states and brain regions, and are far from effective research sharing. This paper proposes a research sharing-oriented approach for functional neuroimaging named entity recognition. The nine most representative entity categories were defined by analyzing the characteristics of task-based functional neuroimaging researches, and a multi-category named entity recognition method was designed based on BiLSTM-CNN. An experiment was performed on literatures obtained from the journal PLoS One. The experimental results show that the precision and recall rates of the proposed method can reach 94.50% and 95.56%, and are obviously superior to existing methods of functional neuroimaing named entity recognition.