S-Space College of Social Sciences (사회과학대학) Dept. of Communication (언론정보학과) Theses (Ph.D. / Sc.D._언론정보학과)
An Algorithmic Approach to Personalized and Interactive News Generation
알고리즘에 기반한 개인화되고 상호작용적인 뉴스 생성에 관한 연구
- Dongwhan Kim
- 사회과학대학 언론정보학과
- Issue Date
- 서울대학교 대학원
- Algorithm Journalism; Robot Journalism; News; Framework; Consolidated Database; Personalization; Interactivity; News System; HCI; Interface; Interaction; Design
- 학위논문 (박사)-- 서울대학교 대학원 : 언론정보학과, 2017. 2. 이준환.
- Algorithms are increasingly playing an important role in the production of news content with growing computational capacity. Moreover, the use of the algorithm is taking up traditional human roles as increasing number of journalistic activities are mediated by software. For instance, the Los Angele Times runs software called Quakebot, which makes automated decisions on publishing news articles on abnormal seismic events. The Associated Press and Forbes have long been publishing algorithm-generated news content in collaboration with narrative-generation algorithm developers since 2014. The Washington Post also joined the trend by developing news reporting software for 2016 Rio Olympics.
We were motivated by the advent of various algorithm-generated news products. We reviewed current practices of algorithm-generated news and classified common algorithmic attributes to derive insights on how to maximize the capacity of the algorithm for more engaging and appealing news content generation. The key opportunity areas we found were 1) broadening depth and breadth of input data enriches algorithmic computation, 2) personalizing the narrative in the context of news readers raises interest, 3) presenting interactive user interface components helps to engage news readers and make them more active news consumers.
We designed an algorithmic framework based on the proposed key concepts and implemented a news generation system called PINGS, which is capable of generating more personalized and interactive news stories. In this thesis, we describe the design process and implementation details that shaped the PINGS. We present a study on how news readers perceive the news values of the content generated by PINGS as well as the comments and opinions on its potential influence in the field and usability and usefulness of the system by recruiting experts for qualitative review. This thesis includes discussions on our approach to design and implement personalization and interactivity functions into a news system, and contributions it makes to the fields of journalism and HCI.