• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04b
• /
• pp.646-648
• /
• 2004

기계학습의 군집화(clustering) 기법은 예제들 간의 유사성에 근거하여 주어진 예제들을 무리 짓는 방법이다. 준감독(semi-supervised) 군집화는 카테고리가 부여된(labeled) 소수의 예제들을 적극적으로 활용하여 군집형태가 보다 자연스럽게 형성되도록 유도하는 군집화 방법이다. 준감독 군집화 문제에서 예제에 카테고리를 부여하는 작업은 현실적으로 극히 제한적이거나 카테고리를 부여하는데 소요되는 비용이 상당하므로, 제한된 자원 내에서 군집화에 효용성이 높을 예제들을 선정하여 카테고리를 부여하는 것이 필요하다. 본 논문에서는 기존 연구에서 능동적 학습의 초기 훈련예제 선정을 위해 제안된 군집기반 훈련예제 선정 방법을 준감독 군집화에 적용하여 군집 결과의 질을 향상시키고자 한다. 군집화를 이용한 예제 선정 방법은 유사한 예제들은 동일한 카테고리에 속할 가능성이 높다는 가정하에 전체 예제를 활용하여 선정하고자 하는 예제 수만큼 군집을 생성 한 후. 각 군집의 중심점에 가장 가까운 예제들을 대표 예제로 선정하여 훈련 집합을 구성하는 방법이다 본 논문에서는 문서를 대상으로 하는 준감독 군집화 실험을 통해, 카테고리를 부여할 예제를 임의로 선정한 경우에 비해 군집화를 이용한 훈련 예제들로 준감독 군집화를 수행한 경우가 보다 좋은 군집을 형성함을 확인하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.05a
• /
• pp.130-131
• /
• 2014

시맨틱 웹은 1998년 웹의 창시자인 Tim Berners Lee가 제안을 한 웹 기술로써 인간의 언어를 이해하고 인간과 쉽게 의사소통이 가능한 네트워크를 말한다. 즉 컴퓨터 스스로 웹에 연결된 정보의 의미를 인식하고 사용자가 필요로 하는 정보를 검색하며 검색된 정보에서 지식을 유추할 수 있는 기능을 제공하는 지능형 웹 환경이다. 이런 시맨틱 웹 개념을 적용한 사례로는 네이버 시맨틱 웹 영화검색 이있다. 본 논문에서는 네이버 시맨틱 웹 영화검색 시스템을 벤치마킹한 도서검색 서비스 설계를 제안하고자 한다. 본 도서검색 서비스는 온톨로지 개념을 적용하여 도서와 관련된 검색 카테고리를 설정하며, 간단한 시나리오는 다음과 같다. 한 권의 책을 검색하면 해당 책과 연관된 첫 번째 카테고리로 출판사, 제작한 년도, 저자, 장르, 관련 검색 도서 등의 데이터들이 상단에 제시된다. 제시된 카테고리에서 임의의 항목을 선택하면 그 하단 공백에 선택된 항목과 연관된 카테고리에 해당하는 항목들이 제시된다. 예를 들어, 출판사를 선택한다면 해당 출판사에서 출간된 도서들이 하단 공백에 열거가 되고 상단 두 번째 카테고리에 원작국가, 저자 관련 책, 수상정보, 공동 집필자, 책을 원작으로 확장된 컨텐츠 등 또 다른 카테고리가 우측으로 생성이 되며 선택을 할 수 있게 된다. 본 논문에서 제안하는 시맨틱 웹 기반 도서검색 서비스는 사용자가 검색하고자 하는 정보를 보다 효율적이고 사용자 중심에서 제공할 수 있다고 사료된다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.5
• /
• pp.470-476
• /
• 2014

A point of interest is a specific point location such as a cafe, a gallery, a shop, or a park. It consists of a name, a category, a location, and so on. Its information is necessary for location-based application, above all category is basic information. However, category information should be automatically gathered because it costs high to gather it manually. In this paper, we propose a novel method to estimate category of POIs automatically using an inner word and local context. An inner word is a word that contains POI's name. Their name sometimes expose category information. Thus, their name is used as inner word information in estimating category of POIs. Local context information means words around a POI's name in a document that mentioned the name. The context include information to estimate category. The evaluation of the proposed method is performed on two data sets. According to the experimental results, proposed model using combination inner word and local context show higher accuracy than that of model using each.

• Journal of Korean Society of Transportation
• /
• v.22 no.3 s.74
• /
• pp.49-58
• /
• 2004

The first and perhaps the most critical and perhaps the most important step in the process of predicting future traffic volume in a region (Zone) is to estimate the number of trips generated in from each traffic analysis zone. Most trip generation models for urban transportation planning, and highway in Korea are regression models. In Korea the category analysis has not been tried for last decades since the proper data such as the household travel behavior data have not been collected. Recently, the comprehensive household travel behavior survey such as ${\ulcorner}$1996 The Household Travel Behavior Survey${\lrcorner}$, ${\ulcorner}$2002 The Household Travel Behavior Survey${\lrcorner}$ has been done. In this paper, the cross-classification tables of Seoul Metropolitan Area including the City of Seoul and Kyonggi Province are estimated by the category analysis. The tables are compared with regression models and ${\ulcorner}$2002 The Household Travel Behavior Survey${\lrcorner}$ data in terms of predictive capabilities in Seoul Metropolitan Area. Improvement strategies for trip generation forecast in Seoul Metropolitan Area are proposed.

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.54-61
• /
• 2016

According to many smart devices are development, SNS(Social Network Service) users are getting higher that is possible for real-time communicating, information sharing without limitations in distance and space. Nowadays, SNS users that based on communication and relationships, are getting uses SNS for information sharing. In this paper, we used the SNS posts for users to extract the category and information provider, how to following of recommend method. Particularly, this paper focuses on classifying the words in the text of the posts and measures the frequency using Inception-v3 model, which is one of the machine learning technique -CNN(Convolutional Neural Network) we classified image word. By classifying the category of a word in a text and image, that based on DMOZ to build the information provider DB. Comparing user categories classified in categories and posts from information provider DB. If the category is matched by measuring the degree of similarity to the information providers is classified in the category, we suggest that how to recommend method of the most similar information providers account.

• Journal of Environmental Policy
• /
• v.6 no.2
• /
• pp.71-96
• /
• 2007

This study aimed at effectively applying the IUCN category system to the protected areas in Korea. In addition, in order to change IUCN category of national parks to IUCN category II and to review the application of IUCN category classification key, a case study was conducted in Sobaeksan National Park. In order to apply the IUCN category system to the protected areas in Korea, a flexible approach appropriate to characteristics in Korea is required for management objectives of protected areas, including protection of wildemess, sustainable use of resources and preservation of cultural and traditional features. In addition, considerations of restrictions on use area and use districts, relative comparison of use types (visit, use of resources, residence) by IUCN category and use of combined classifications are necessary. Principles for the application of the IUCN category include (1)exclusion of wilderness protected areas (Ib), (2) extremely limited use regarding the sustainable use of natural resources(sum of natural preservation area and natural environment area is over 95%), (3) considerations of management conditions, including residential occupation level, (4) preservation of ecosystem services, and (5) use of combined classifications. In addition, in accordance with these principles, IUCN category classification key was suggested. When this was applied to the case study area, Sobaeksan National Park was classified as IUCN Category II and Taxus cuspidata community, which is designated as a natural monument, was classified to be Category Ia. Classification key suggested in this study may be used as basic data for applying categories in the future. Since detailed review on the practical improvement direction of laws and regulations and systematic alternatives, which are required before introducing IUCN category, are poor, in order to manage the protected areas efficiently by applying the IUCN category in the future, studies on management means appropriate to the conservation objectives of each category are necessary. This would allow management differentiated for each category.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.10a
• /
• pp.3-6
• /
• 2012

Due to recent rapid growth of mobile market, the modern people increasing make use of mobile contents as a means to obtain the desired information quickly by overcoming various restraints of a computer. The wide range of recommended contents, however, takes much time in selection of contents. To resolve such issues, a system that predicts the contents desired by the user and makes an accurate recommendation is necessary. In this paper, in order to provide the desired contents in line with the user demands, a method to increase select the number of recommendation using cooperative filtering is proposed. In the first step, the categories are formulated with super-classes and the similarity between the target customer and users is found, and the nearest-neighbors are constituted to find the preference predictions between super-classes, and the super-class with the highest resulting value is recommended to the target customer. In the second step, the preference predictions between sub-classes are found and the sub-class with the highest value is recommended to the target customer. In the experiment, mobile contents are recommended through super-class-based cooperative filtering, and then the mobile contents are recommended through sub-class-based cooperative filtering, and sub-class collaborative filtering method to select a high number of verification.

• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.562-566
• /
• 2009

In this paper, we present a novel system to categorize web image search results using emotional concepts and to browse the results more conveniently and easily. The proposed system can categorize search results into 8 emotional categories based on emotion vector, which obtained by color and pattern features. Here, we use Kobayashi’s emotional categories: {romantic, natural, casual, elegant, chic, classic, dandy and modern}. With search results for a given query, the proposed system can provide categorized images for each emotional category. With 1,000 Yahoo! search images, we compared the proposed method with Yahoo! image search engine in respect of satisfaction, efficiency, convenience and relevance with a user study. Our experimental results show the effectiveness of the proposed method.

• Journal of KIISE
• /
• v.42 no.5
• /
• pp.664-670
• /
• 2015

With the rapid progression of the Internet and smartphones, YouTube has grown significantly as a social media sharing site and has become popular all around the world. As users share videos through YouTube, social data are created and users look for video recommendations related to their interests. In this paper, we extract users' social category based on their social relationship and social category classification list using YouTube data. We propose the YouTube recommendation algorithm using the extracted users' social category for more accurate and meaningful recommendations. We show experiment results of its validation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2014.11a
• /
• pp.737-740
• /
• 2014

최근 SNS 사용자가 급증하면서 매우 다양하고 방대한 양의 글이 여러 종류의 SNS를 통해 생성되고 있다. 그중 트위터는 정보의 전달 및 확산에 상당히 유용한 도구로 사용되고 있다. 이러한 트위터의 사용자 트윗은 뉴스, 음악, 사진, 여행 등 다양한 형태로 등장한다. 또한 트위터는 해시태그라는 사용자 정의 태그를 사용하는데 이는 트윗의 키워드 및 핵심을 쉽게 표현할 수 있도록 해주는 효과적인 수단이다. 최근 상당히 많은 양의 트윗의 생성에도 불구하고 이를 다양한 카테고리별로 분류할 수 있는 연구가 많이 진행되지 않았다. 따라서 본 논문에서는 해시태그를 이용해 트윗의 핵심을 파악하고 수많은 트윗을 다양한 토픽별로 분류할 수 있는 기법을 제안한다. 우선 다양한 카테고리의 인기 해시태그가 포함된 트윗을 수집하고 수집한 트윗에서 해시태그별 키워드를 추출한다. 그리고 코사인 유사도를 통해 해시태그별 내용 유사도를 파악하여 각 카테고리 내의 해시태그가 얼마나 유사한 내용을 지니고 있는지 파악한다. 마지막으로 사용자 트윗이 입력되면 모든 카테고리와 유사도를 비교하여 가장 유사도가 높은 카테고리를 찾아 추천해준다. 제안된 기법을 바탕으로 프로토타입을 구현하고 실험을 통해 성능을 평가한다.