We developed a suggestion system that recommends related Korean Medicine symptoms from natural language description of patient's health condition. First, we collected medical information from 200 patients who suffer digestive disorder. Then, three groups of Korean medicine doctors belonging to each of the major Korean medicine associations examined clinical report forms (CRFs) of the patients and specified primary symptoms of the patients. The Korean medicine doctors also attached text evidences of their decision based on the descriptions in the CRFs. Those text descriptions and the symptoms were used as the input text and the label of the system. The features of the multi label support vector machine are term frequencies of the input text. Those terms are the morphemes from Korean morphological analyzer. The model performance measure was the area under cover (AUC) values that is calculated based on the predicted ranks of the true symptom. All AUC values of the models from three independent datasets generated by the three different groups of Korean medicine doctors are more than 0.85 which suggests our system shows consistent and robust performances regardless of the given datasets. Our study is an artificial intelligence (AI) application to Korean medicine and our approach is useful to explore proper symptom names from the thousands of Korean medicine terms. It can be expanded to application to other kinds of disorders easily. Better performance can be achieved by adapting advanced text processing technique and AI models.