DSpace Community: KAIST Graduate School of Web Science Technology
http://hdl.handle.net/10203/175371
KAIST Graduate School of Web Science Technology2024-03-19T01:24:26Z교차 언어 전이를 이용한 한국어 프레임 의미 분석
http://hdl.handle.net/10203/295570
Title: 교차 언어 전이를 이용한 한국어 프레임 의미 분석
Authors: 함영균
Abstract: 본 논문에서는 한국어 의미 분석을 위한 교차 언어 전이 방법론을 다루었다. 프레임 의미 분석이란 프레임넷의 프레임 의미론에 기반하여 텍스트의 의미를 이해하는 자연언어처리 태스크중 하나이다. 최근 다국어 프레임 의미 분석에 대한 관심이 높아지고 있으나, 비 영어권 프레임넷 구축이나 의미 분석기 개발은 대상언어 데이터 및 모델 구축의 비용과 복잡함에 의해 도전과제로 여겨지고 있다. 본 논문에서는 교차 언어 전이 방법이 기존의 풍부한 언어자원을 가진 언어(예: 영어)의 지식이 효과적으로 대상 언어의 프레임 의미 분석에 적용될 수 있음을 주장한다.
본 논문은 한국어 프레임 의미 분석기을 위해 다음의 세 가지 방법을 적용했다. 첫 째로 주석 전이 방법을 통해 기존에 존재하는 프레임넷 주석으로부터 한국어 프레임넷의 초기 주석을 전이하였다. 두 번째는 교차 언어 전이 학습 방법이다. 언어 독립적 아키텍쳐를 통해 영어 프레임넷의 활용으로 한국어 모델의 성능을 향상시켰다. 세 번째는 도메인 적응이다. 한국어의 다양한 도메인에 적용될 수 있는 모델을 위하여 비지도학습방법을 통해 도메인 적응력을 보완하였다. 상기의 방법들에 대한 정량적 평가와 정성적 평가를 통하여, 본 논문은 교차 언어 전이 방법을 통한 한국어 프레임넷과 프레임 의미 분석기 구축이 효과적인 방법임을 보였다.
Description: 학위논문(박사) - 한국과학기술원 : 웹사이언스대학원, 2021.2,[iii, 103 p. :]2021-01-01T00:00:00ZIdentification and analysis of key factors driving longevity in online social platforms
http://hdl.handle.net/10203/265073
Title: Identification and analysis of key factors driving longevity in online social platforms
Authors: Park, Kunwoo
Abstract: Having a substantial number of loyal users is a key for the success of online social platforms. When building a social platform where which users enjoy over a long period, it is crucial to identify predictive signals for long-term engagement, also known as user longevity. This dissertation analyzes and identifies key factors driving user longevity from three data-driven case studies.First, in an online multiplayer game where rich user behaviors are digitally logged, I investigated how factors driving longevity vary as user level increases. While achievement was the main factor in the lower-level stages, social features became the most important in predicting long-term engagement even after reaching the highest level. The varying key indicators across different levels suggest that it is necessary to consider virtual life phases to analyze user longevity indicators precisely. Second, I analyzed shared logs of MyFitnessPal on Twitter via the social sharing mechanism, and investigated whether additional information from another platform can be predictive of user longevity in the target platform. Cross-platform analysis demonstrates that features extracted from Twitter are predictive of long-term engagement in MyFitnessPal. This study shows that utilizing a supplementary social network via social sharing mechanism enables us to investigate potential factors driving longevity more diversely. Third, from a case study on the Reddit science community, I analyze the effects of disclosing offline social status on continued usages over an extended period. The design mechanism that reveals offline social status has been adopted to complement online reputation systems; however, it is not fully investigated how the design affects future engagement in online social platforms yet. From causal inferences on user longevity in Reddit science community, I found that disclosing academic degrees has a positive effect on continued usage over an extended period in the community. On the other hand, the design mechanism decreased social interactions toward those who do not reveal their offline status. The results suggest that the design mechanism should be carefully introduced to online social platforms because disclosed offline social status can create mixed effects on future engagements in online social platforms.To summarize, this dissertation conducts three data-driven case studies on identifying key factors driving user longevity in online social platforms, and presents a common finding on the importance of social factors across the three studies. Although each finding is from a specific case and hence cannot be generalized, future cross-platform studies could lead to building a social platform that users enjoy over an extended period.
Description: 학위논문(박사) - 한국과학기술원 : 웹사이언스대학원, 2018.2,[iv, 57 p. :]2018-01-01T00:00:00Z(A) wearable-based mobile system for managing itching conditions
http://hdl.handle.net/10203/265072
Title: (A) wearable-based mobile system for managing itching conditions
Authors: Lee, Jong In
Abstract: Severe itching conditions such as eczema or atopic dermatitis can have a significant impact on one’s quality of life. Unfortunately, many of these conditions cannot be cured, and the focus is often on properly controlling or managing the condition. Thus, it is important to understand or objectively monitor how one’s scratching behavior changes, based on medication or treatment or environmental conditions. In this work, we explore how wearable devices can support people with itching conditions to better manage their conditions. We carried out a three-phase study with 40 participants and 2 dermatologists to understand the implications of various system features and designs. Based on interviews with patients and doctors, we incorporated medical guidelines for treatment and patients’ needs in the proposed Itchtector – a smartwatch-based mobile system to monitor itching behaviors and provide objective information about the user’s scratching behaviors. Using the Itchtector prototype, we evaluated performance and possible acceptance with subjects.
Description: 학위논문(박사) - 한국과학기술원 : 웹사이언스대학원, 2018.8,[iv, 51 p. :]2018-01-01T00:00:00ZIdentification and assessment of user-generated advertisements on social network services
http://hdl.handle.net/10203/241968
Title: Identification and assessment of user-generated advertisements on social network services
Authors: Park, Jaimie Yejean; 박예진
Abstract: With the rapid development of social media, we have witnessed a radical change in the advertising landscape. Advertising is no longer just a one-way communication; advertising on social media is a collaborative and iterative process which oftentimes involves SNS users in creating, modifying, and propagating the advertising content. In this thesis, we use mixed-method approach to address three important questions on user-generated advertising: 1) How do we identify user-generated advertisements on social media? 2) How do users value different user-generated advertising strategies on social media? 3) How does the audience react to user-generated advertisements?
Description: 학위논문(박사) - 한국과학기술원 : 웹사이언스대학원, 2017.8,[iv, 49 p. :]2017-01-01T00:00:00Z