Accuracy of the Sentence-BERT Semantic Search System for a Japanese Database of Closed Medical Malpractice Claims

In this study, we developed a similar text retrieval system using Sentence-BERT (SBERT) for our database of closed medical malpractice claims and investigated its accuracy. We assigned each case in the short Japanese summary accident as well two labels: category was classified hospital department mainly, process indicated failed procedure. evaluated accuracies with labels three different multilabel evaluation metrics. For encoders SBERT, employed pretrained BERT models, UTH-BERT NICT-BERT, that were trained on huge corpora, performed iterative optimization to train SBERTs. The systems SBERTs more than 15 points higher those Okapi BM25 SBERT system.