| Abstract |
In my previous talk, I introduced the purpose and design of the SWiFT game. In this talk, I will present what we have learned from the first time the SWiFT game was actually played with two ferocious teams. A formal abstract follows:
The Semantic Web (SW) holds the promise of improving computer-supported collaboration. It does so by increasing the transparency and reusability of representation of information, specifically in a computational sense. We argue that the SW will strongly benefit from an improved human capability of creating high quality SW representations on the fly. In that context we pursue people to become fluent in ‘SW writing’ during the process of creating information for a specific purpose. Using the Web as a medium to share this information, these thoughts are available immediately for others to reuse and extend, ideally creating a process of ‘collective thinking’. The application context happens to be that of food research, but we believe that our experiences are generic and extend to other areas of science and to collaboration in general as well.
On the Web, direct expression of thoughts is typically seen in blogs. These are entirely set in natural language and do not meet our requirements in terms of creating a computational (formal) representation. (Normal) Wikis, media for informal collaboration, are also used for recording thoughts, but have the same disadvantage. Semantic Wikis are a step toward ‘SW writing’, however, they typically formalise only part of the information, avoiding modelling problems that appear in full text translations. Moreover, they do not stress approaching fluency.
To obtain some understanding of the oddities and pitfalls of ‘real-time’ SW writing we developed a format for knowledge representation (essentially based on RDF(S)), a set of instructions (a constitution) and a supporting tool. Moreover, we present these to a test set of writers in the form of a game, called SWiFT. In this game, competing teams translate the same natural language text into a formal representation, each player covering a fragment that doesn't overlap with others of his team. Then teams challenge each other's full translations by posing questions that have a very specific answer based on the content of the text (`question attack'). Each team then tries to construct a formal query that derives the answer from their translation (`algorithmic defense'). A short translation time, a low average complexity of queries, and correct answers contribute positively to the final score of a team. In this talk we present our observations on the game played by two teams of writers. It appears that this type of real-time modelling uncovers typical modelling issues that are not seen in other conditions. |