Building a controlled health vocabulary in Japanese

This study is aimed at developing a controlled clinical vocabulary for use in electronic patient record (EPR) systems.

In this paper, we propose a model for building the vocabulary. The model is composed of a Canonical Term Dictionary, an Atom Dictionary, a Composite Atom Dictionary, and an Index. Parsing and composing functions are included in this model. Canonical terms were extracted from reference terminologies. Atoms were extracted from the Canonical Term Dictionary and reduced to a set from which the Composite Atom Dictionary can be built. The index was built to link these two dictionaries. For testing the model, we compiled a sample vocabulary and applied the model to a SNOMED translation system (English to Japanese) and a term similarity estimation system.

The sample vocabulary consisted of 15,600 atomic terms and 4,450 composite terms. 33,441 SNOMED terms were translated by the SNOMED translation system. The system gave adequate Japanese candidates in 56.3% of cases. The similarity estimation system found an average of 5.4 candidates when the equality ratio was over 50%.

The trial applications produced good results. The model seems promising for building a standard clinical vocabulary system. This system can be applied in certain other Asian countries, such as China and Korea.