As the first point of contact for patients, General Practitioner (GP) plays an important role in the National Health System. An accurate primary diagnosis from the GP on the patient will relieve specialists’ pressure and save time from confirming the patient’s condition and doing examinations. Because GP has broader but less specialized knowledge, the accuracy of their diagnosis is limited. Therefore, it is imperative to introduce an intelligent system to assist GP to make decisions. This paper proposed a hybrid architecture, which fuses the features of words from different representation spaces. Two data augmentation methods (Complaint-Symptoms Integration method and Symptom Dot Separating Method) have been proposed to integrate essential information into the training data. Experiments demonstrate that this hybrid architecture has good performance in the classification of 4 common neurological diseases 1. Finally, this paper develops an AI diagnosis assistant webapp which leverages the superior performance of this architecture to help GPs complete primary diagnosis efficiently and accurately.