Project: Authoring Realistic Learning Environments with Stories (ARLES)
A project of the USC Institute for Creative Technologies, 11/1/2011 to 10/30/2013.
Many classroom instructors and tutors draw from their own experiences when training complex cognitive skills, telling stories of their own successes and failures to illustrate specific points or to ensure memorable impact. In immersive training environments for the military, elements of stories of real-world experience are interwoven into fictional training scenarios designed to give trainees memorable (simulated) experiences of their own. To succeed in the authoring of effective scenarios, training developers need easy access to large repositories of nonfiction narratives told by real-world practitioners.
Can we automate the collection and organization of real-world narratives in order to support the development of immersive training environments?
Our research approach has been to create the ICT Story Pipeline, a software architecture for collecting, analyzing, retrieving, and utilizing millions of personal narratives posted to public weblogs.
- Each day, we process the text of roughly 1.5 million English-language weblog entries posted to public weblogs.
- Using statistical text classification techniques, we identify the small percentage of personal narratives that appear in weblog posts (tens of thousands each day).
- We add each personal narrative to our story database repository, now tens of millions large, and index them for retrieval using a contemporary text retrieval engine.
- We search our collection using two novel search tools, StoryUpgrade and PhotoFall, which utilize relevance feedback to induce statistical topic model of stories related to specific activities.
- We identify hundreds of civilian activities that are analogous to US Army tasks, and have used our search tools to build collections of personal narratives for each of them.
- We provide software interfaces that enable training developers to link stories directly to training objectives, which can then be provided to trainees in customized after-action reviews.
- We empirically investigate the usefullness of civilian narratives to the articulation of US Army learning objectives in controlled experiments.
- We provide training developers access to civilian story collections through a novel narrative-based collection browser.
- We investigate the practical and ethical considerations of using personal stories in public weblogs through first-person interviews with prolific bloggers.
This project was conducted by Andrew S. Gordon (PI), and
Christopher Wienberg, with contributions from
H. Chad Lane,
Jessa Sininger, and
- Wienberg, C., Roemmele, M., and Gordon, A. (2013) Content-Based Similarity Measures of Weblog Authors. The 4th Annual ACM Web Science Conference (WebSci-13), May 2-4, 2013, Paris, France.
- Wienberg, C. and Gordon, A. (2012) PhotoFall: Discovering Weblog Stories Through Photographs. Proceedings of The 21st ACM Conference on Information and Knowledge Management (CIKM-2012), October 29-November 2, 2012, Maui, HI.
- Gordon, A., Wienberg, C., and Sood, S. (2012) Different Strokes of Different Folks: Searching for Health Narratives in Weblogs. 2012 ASE/IEEE International Conference on Social Computing, September 3-5, 2012, Amsterdam, The Netherlands.
- Campbell, A., Wienberg, C., and Gordon, A. (2012) Collecting Relevance Feedback on Titles and Photographs in Weblog Posts. Proceedings of the 2012 International Conference on Intelligent User Interfaces (IUI-12), February 14-17, 2012, Lisbon, Portugal.
This project was conducted as part of the mission funding of the USC Institute for Creative Technologies, contract number #W911NF-04-D-0005 (Task 28), project period of performance 11/1/2011 to 10/30/2013.
Andrew S. Gordon
Institute for Creative Technologies
University of Southern California
12015 Waterfront Drive
Los Angeles, CA 90094-2536
Phone: (310) 574-5700
Fax: (310) 574-5725