The medical dictionary that can be used by AI systems, which was the challenge of the AI Hospital Project, was constructed under the guidance of Dr. Aramaki of Nara Institute of Science and Technology. This dictionary is characterized by its format as a network-type dictionary. Broadly speaking, it consists of two structures: a “relation table based on weighting and association between terms” and a ”term table listing a variety of medical terms”. The total number of medical terms is 380,000. Automatic voice input system makes it possible to generate " natural conversational sentences" by collecting corpus. However, since conversations in medical settings involve differences, special phrasing such as dialects may exist depending on the environment in which the term appears. Therefore, translation using a dictionary is required. The relationship table, which is a major feature of this dictionary, is a mechanism that allows you to visualize the frequency of co-occurrence between terms by structuring the relationships between terms, such as drugs and diagnosis names, symptoms and diagnosis names, etc. is being built. The terminology table consists of "disease name/symptoms", "medicine", etc., and also supports pronunciation, linking with ICD codes, and English translation.
These tables are intended for use in "AI diagnosis candidate support systems," "automatic voice input support systems," and "hospitals, clinics, etc."
In actual verification research, it has been confirmed that IBM's automatic voice input system and diagnostic support system have improved performance. We hope that we can contribute to ``medical care that meets the needs of our patients.''