Content

Arkose (Ph.D. thesis)

ARKOSE: A Prototype Mechanism and Tool for Collaborative Information Generation and Distillation

Abstract:

The goals of this thesis have been to gain a better understanding of collaborative knowledge sharing and distilling and to build a prototype collaborative system that supports flexible knowledge generation and distillation. To reach these goals, I have conducted two user studies and built two systems.

The first system, Arkose 1.0, is a prototype collaborative distillation system for a discussion space, which provides a set of augmentative tools to facilitate the filtering, structuring, and organizing of discussion information. Arkose 1.0 supports editors to distill a discussion space incrementally and collaboratively, and allows a gradual increase in the order and reusability of the information space.

The study of an online question-answering community, Naver Knowledge-iN, investigates users’ knowledge sharing behaviors in a large online question-answering community. Through the analyses of a large quantity of question/answer pairs and 26 user interviews, the study analyzes the characteristics of knowledge generation and user participation behavior and gains insights into their motivations, roles, usage and expertise. It reveals that the limiting nature of the reply interfaces of Knowledge-iN leads to the accumulation of simple and easy questions and answers. This tendency is encouraged by the point system that rewards users who answer many questions quickly.

Arkose2 is designed and implemented based on the lessons and insights gained in building Arkose 1.0 and examining Naver Knowledge-iN. Arkose2 provides a host of additional interaction mechanisms and supportive tools over Arkose 1.0 that assists users to flexibly generate knowledge and distill and organize it better.  Finally, the evaluation of Arkose2 reveals a number of insights, issues, and lessons about users’ distillation activities of discussion spaces and features of Arkose2.

These together provide valuable lessons and insights for the architecture and features of the next generation collective intelligent system.

The following shows some of Arkose2′ features that augment human editors’ collaborative distillation tasks of a large online discussion space.  Arkose2 visualizes a discussion space and provides interactive and supportive tools, and meta-information collected by the underlying text-analytic engine.

—————————–

Visual Navigator

A partial view of Visual Navigator. It supports discussion discourse, summarization scaffolding, editing, and a variety of progress indicators. Raw discussion data, user interactions through comments, summaries, keywords, and other meta-information co-exist in the space.

—————————–

Dividing a complex question into sub-questions

Handling a complex question. The interfaces allow breaking a question or topic into subtopics. This allow a more organized discussion.

—————————–

Automatic clustering of posts into similar topics

Automatic clustering of the posts through the underlying text-analytic engine. Posts of the same color indicate they are likely about similar topics.

—————————–

Proportional tagging system for better representing a content

An experimental tagging system. Instead of just adding tags, a user can quickly set the weight of each tag by dragging the handle in the bar indication the importance of each tag.

—————————–

An actual user’s distillation result.  Most posts were suammarized into a few summary objects.

A user’s distillation progress. Having a topic scaffolding helped a user to create a more organized and meaningful summaries in general.