한의학 증상 질의 확장을 위한 단어 임베딩 기반의 클러스터링 기법
Word Embedding Based Clustering Method for Query Expansion of Korean Medicine Symptoms
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
정보화 시대에는 막대하게 축적된 정보량으로 인해서 스마트한 정보 검색이 필수적이다. 이를 지원하기 위해서 질의 확장을 통해서 사용자의 정확한 검색을 지원하고 있다. 본 연구에서는 한의 진료 지원 시스템의 질의 확장을 위해서 최초 질의로부터 생성된 처방 정보에서 증상명을 추출하여 벡터공간으로 임베딩 및 클러스터링하고 각 클러스터의 품질을 평가하여 높은 순위의 증상명부터 확장 질의로 추천하는 방법을 제안하고 한의사에 의한 정성평가 및 검색 결에 대한 정량평가를 실시하였다. 실험에서는 연관성, 포괄성, 유용성의 세 가지 지표를 평가하였으며, 세 가지 지표에 대한 평가 결과를 통합적으로 분석하여, 제안된 모델이 한의 진료 지원 시스템의 연관 검색어 제시를 위한 질의 확장모델로 가장 적합한 것으로 드러났다. 향후 사용자 검색 기록이 수집된다면, 더욱 개선된 질의 확장을 통한 연관 검색어 제공이 가능해질 것으로 기대된다.
Abstract
In the information age, smart search is essential due to the enormous amount of accumulated information. However most of the query entered into search engines are general nouns with various meanings and are only about 2.4 words on average, making it difficult for search engines to grasp the exact search intention of users. To mitigate this situation, figuring out exact search intention of users is supported by query expansion. In order to develop the query expansion of Korean Medicine clinical decision support system (KM-CDSS), we suggest novel algorithm that consist of 4 steps; first, symptom names are extracted from the prescription information fetched by the initial query. second, extracted symptoms are embedded into a vector space. third, vector space are clustered and each cluster’s quality is evaluated by silhouette coefficient. finally, each words which are laid nearest to the center is suggested as representative symptom. In the experiments, we examined relevance and comprehensiveness qualitatively by KM doctors and utility are analyzed quantitatively. The results of the evaluation of the three indicators were analyzed in an integrated manner. The proposed model showed good enough results in relevance and comprehensiveness test and best result in utility test. It turned out that the proposed model is most suitable as a query expansion model for KM-CDSS. If user search logs are collected in the future, it is expected that related search words will be provided in more sophisticated ways through improved query expansion.
참고문헌 (19)
- P. Suthira, "A query suggestion method combining TF-IDF and Jaccard Coefficient for interactive web search", Artificial Intelligence Research, 4(2), pp.119~125 (2015)
- C. Fei, "Query auto completion in information retrieval", Universiteit van Amsterdam, ~ (2016)
- 곽창욱, "유사 단어 커뮤니티 기반의 질의 확장", 정보과학회논문지, 41(12), pp.1058~1065 (2014), 한국정보과학회
- A. H. Kumar, "Query expansion techniques for information retrieval: a survey", Information Processing & Management, 56(5), pp.1698~1735 (2019)
- O. Jessie, "A survey of query expansion, query suggestion and query refinement techniques", 2015 4th International Conference on Software Engineering and Computer Systems (ICSECS) (2015), IEEE
- H-C. Jang, "Ontology for medicinal materials based on traditional Korean medicine", Bioinformatics, 26(18), pp.2359~2360 (2010)
- H-C. Jang, "User guiding information supporting application for clinical procedure in traditional medicine", pp.2013, HIMI
- Y. Xu, "Query dependent pseudo-relevance feedback based on wikipedia", Proc. of the 32nd SIGIR, pp.59~66 (2009)
- Z. Liu, "Query expansion based on clustered results", Proc. of the 37th VLDB, 4(6), pp.350~361 (2011)
- O. Vechtomova, "Query expansion with long-span collocates", Information Retrieval, 6(2), pp.251~273 (2003)
- K-C. Lee, "Sentence embeddings using korean corpus" (2019), acornPub
- P. Bojanowski, "Enriching word vectors with subword information", Transactions of the Association for Computational Linguistics, 5, pp.135~146 (2017)
- 박사준, "의미 기반 문단 분할을 활용한 문단 기반 K-Means 클러스터링", 한국지식정보기술학회 논문지, 12(1), pp.157~164 (2017), 한국지식정보기술학회[DOI]
- 장현성, "Data-mining Based Anomaly Detection in Document Management System", 한국지식정보기술학회 논문지, 10(4), pp.465~473 (2015), 한국지식정보기술학회
- S. David, "Web-scale k-means clustering", Proceedings of the 19th international conference on World wide web (2010)
- S. Artur, "Performance evaluation of the silhouette index", International Conference on Artificial Intelligence and Soft Computing (2015), Springer
- "Publication Committee of Yearbook of Traditional Korean Medicine", 2018Yearbook of Traditional Korean Medicine, ~ (2020)
- G. Koutrika, "Data clouds: summarizing keyword search results over structured data", Proc. of EDBT, pp.391~402 (2009)
- M. Zhongrui, "New assessment criteria for query suggestion", Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval, pp.1109~1110 (2012)