Center for Creation, Content and Technology (CCCT) Seminar
Friday 17 January 2014, 16.00-17.00 (followed by drinks), Science Park 904, room C1.112.
Under the CCCT umbrella, researchers from the humanities, the social and behavioral sciences, and the natural sciences collaborate in a multidisciplinary setting on information-rich research topics. CCCT is organizing monthly seminars in which speakers report on research activities that are of shared interest.
Speaker: Sally Wyatt (University of Maastricht and KNAW)
Self-reported data in medical and social research
Speaker: Toine Pieters (University of Utrecht and Decartes Centre)
Towards historical text mining in public media: WAHSP/BILAND
The history of science and technology and cultural history –disciplines which have traditionally distanced themselves from computer based research– arguably have the most to gain from digital methodologies. A central feature of both disciplines is the close examination of written, spoken and visual source material as a means of exploring discursive formations and historical patterns of continuity and discontinuity. Tracking cultural and scientific ideas, notions and discussions in the digital age holds the promise of exciting new insights in long-term patterns of economic, scientific and cultural change.
The WAHSP/BILAND research teams so far found that, in terms of methodology, semi-automatic document selection fits rather well with historical research as an alternative to manual browsing or random sampling, facilitating the combination of qualitative and quantitative approaches. Through text mining and visualisation, new insights can be gained from an initial selection. Word clouds based on word counts depicting the linguistic context within which keywords occur are instrumental in helping the historian with expert knowledge of the domain to combine and compare different historical periods in a free associative manner. Each query immediately yields a document selection without laborious sampling. This speeds up the heuristic process considerably. Exploring word associations and metadata, as well as histogram visualisations of the documents over time, can lead to improved queries and therefore to a more representative document selection. Such quantitative analysis enhances the knowledge of the historian. A clear benefit of using exploratory searches is to allow the historian to recycle previous insights to investigate new research questions. Comparing document selections using quantitative analysis helps to validate these selections, making them less arbitrary and thus more representative.
Moderator: Maarten de Rijke (Informatics Institute, UvA)
Science Park, room: C1.112
Science Park 904
1098 XH Amsterdam