발간년도 : [2021]
| 논문정보 |
|
| 논문명(한글) |
[Vol.16, No.4] Ranking of Similar Symptom Expressions in Korean Medicine Based on Chinese Character Similarity and Machine Translation Similarity |
|
| 논문투고자 |
Ho Jang, Sang-Jun Yea, Sang-Hun Lee, Sang-Kyun Kim |
|
| 논문내용 |
Korean medicine ontology is a database containing information of symptoms and prescriptions from various Korean medicine literatures. Although the database covers diverse types of diseases and treatments, it is not easy to search for the information that user wants because there are many different expressions representing similar or same meanings. To facilitate the information retrieval, we studied the ranking methods for suggesting similar symptom expressions based on two different approaches. In the ranking approach based on character similarity, TF-IDF vectors representing occurrences of specific characters were used. Similar symptom expressions with the query symptom expression were suggested in order of high cosine similarities between two vectors. In the ranking approach based on machine translation similarity, all symptom expressions were machine-translated to English expressions. TF-IDF vectors representing occurrences of specific English word tokens were used for calculating similarity. The symptom expressions with high cosine similarities were suggested as having high relevance with the query. We have evaluated usefulness of the ranking method by comparing the lists sorted by each ranking approach and the list of similar symptom expressions manually created by Korean medicine doctors. The Area under cover (AUC) of the performance shows 0.9245 for the approach based on character similarity and 0.5193 for the approach based on machine translation similarity respectively. The AUC value of the combination of two ranking approaches are 0.9354 suggesting the combination of different approaches can improve the performance. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
