http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
조류 인플루엔자 예방 및 면역 증진을 위한 천연 사료 첨가제 특허동향 분석
박찬정(Chanjeong Park) 한국콘텐츠학회 2018 한국콘텐츠학회논문지 Vol.18 No.3
본 연구에서는 조류 인플루엔자 예방 및 면역 증진을 위한 천연 사료 첨가제의 특허동향을 분석하였다. 2017년 12월 31일 이전에 공개된 한국 및 중국 특허를 대상으로 검색하였다. 한국은 53건, 중국은 105건의 관련특허를 유효특허로 선정하였다. 이 중에서 등록된 특허는 한국은 38건, 중국은 18건이었다. 한국과 중국 모두 2000년 후반부터 출원 건수가 증가하였으며, 특히 중국은 2016년의 특허 출원활동이 활발하였다. 한국특허에서 주로 사용된 사료 첨가제는 녹차, 울금, 강황, 진피, 패모, 행인 황련, 중국특허에서는 판람근, 금은화, 연교, 감초, 황기, 황금, 산사 등이 많이 사용되었다. 두 국가 모두 한방재료를 이용한 사료첨가제가 많았다. 한국은 대학, 연구소 및 바이오 관련 기업의 출원이 많았고, 중국은 개인 출원인의 비중이 높았다. This study investigate the patent trend of natural feed additives for avian influenza prevention and immunity enhancement. The search scope is Korean and Chinese patents published before December 31, 2017. As a result, I found 53 Korean patents and 105 Chinese patents. Korea and China has increased the number of patent applications since late 2000. In particular, China filed the most patent applications in 2016. Both Korea and China, the oriental medicine materials were mainly used in patent claims. Korea has many patent applications that are submitted by universities, research institutes and bio companies, but China has a high proportion of individual patent applicants.
박찬정(Chanjeong Park),김기용(Kiyong Kim),성동수(Dongsu Seong),이건배(Keonbae Lee) 한국정보기술학회 2014 한국정보기술학회논문지 Vol.12 No.9
Recently, Big Data is studied very much around the world, used in various ways in many fields. Among them, in order to predict a promising technology and prospect a future industry, research on Big Data using patent documents has been increased. It is possible to analyze the techniques in patent documents by the IPC classification code. As the patents are increased each year, the need for automatic classification of IPC has increased. In this paper, we do a research on IPC automatic classification using machine learning, terms clustered using the intimacy between terms in patent documents. As the results, when using the term clustering, classification rate is improved overall. Classification accuracy is found to be excellent in sections where to apply the application rate of the low feature selection.
박찬정(Chanjeong Park),성동수(Dongsu Seong),이건배(Keonbae Lee) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.4
For the systematic classification, retrieval and efficient management of patent document, all patent documents are being given the IPC code. Classification experts at KIPI classify all patent documents for IPC classification by hand, currently in Korea. In this paper, we show the useful machine learning methods for automatic classification of IPC codes to each patent document. We conduct a performance evaluation of various feature selection algorithms and document classifier models. As learning and test data set, 8 kinds of patent documents extracted from IPC (sub)class are used. As the result, boolean weighting method, chi-square statistics and information gain are efficient for the feature selection, and the feature selection ratio of 20-40% is most efficient in the classification accuracy and computational cost. Also, SVM is slightly higher than the Naive Bayesian classification performance, but there were no significant differences.
박찬정(Chanjeong Park),김기용(Kiyong Kim),성동수(Dongsu Seong) 한국정보기술학회 2014 한국정보기술학회논문지 Vol.12 No.3
Convergence technology is actively being studied globally. Recently, IPC(International Patent Classification) used for analyze convergence technologies, in particular, multiple IPC classified patent documents are used. Thus, using the IPC for the convergence technology to retrieve, to analyze the IPC classified patent documents should be made fast and accurate. In this paper, automatic classification of patent documents included multiple classification of IPC using KNN classifiers, and the methods of selecting the appropriate features and comparing rates are tested. Feature selection methods Chi square statistics, Information gain, Mutual information and Dominant information that is proposed method in this paper were used. As a result, most of the feature selection methods are available feature selection ratio within 10%, performance is good. Especially, the rate of around 4% in the case of Dominant information when used as better than the other feature selection methods were compared with other feature selection method showed that KNN classified speed is faster.