Identification Semantic Text of Indonesian Medical Terms from Question-Answer Data

Searching for health-related information online is becoming more difficult due to the proliferation of multiple-meaning phrases such as semantic words. This study examines the semantic process in medical terms using a collection of doctor's answer texts which requires finding an appropriate model for recognizing text in pairs of comparable Indonesian medical phrases or terminology synonyms. This study contributes to finding an automatic semantic text detection using a word embedding approach to identify text in pairs of similar Indonesian medical terms. We selected 108 pairs of annotated medical terms (Biomedical Named Entity Recognition (Bio-NER)) in Indonesian based on a collection of doctor texts with 60 pairs of similar words and 48 pairs of dissimilar words. Our dataset was processed with the word embedding approach of FastText and BioWordVec. There are two approaches of BioWordVec: with (BioWordVec-2) and without (BioWordVec) translation process. We compared the performance of FastText, BioWordVec, and BioWordVec-2 using measures like accuracy, specificity, and sensitivity. The results show that the BioWordVec-2 model performs better than other models in identifying similar pairs.