Constructing Broad-Coverage Knowledge Repositories by Collecting Knowledge from Volunteers

Опубликовано 6 сентября 2016, 5:21

Availability of broad-coverage knowledge repositories can potentially enable the next generation of artificial intelligence and natural language understanding applications. An emerging approach to constructing such repositories is by collecting them from volunteer contributors. Such mass collaboration approach allows posing finely constrained or extensively contextual questions as well as extensive validation of knowledge. However, it is most effective only if we can figure out how to guide and manage the effort of the contributors. This talks will cover: - empirical data on quality of the knowledge collected in a representative system, - types of knowledge which collection from volunteers should target - methods of collecting more interpretable knowledge, - methods for incenting contributors to give better contributions, and - present a high-level view on management (collection and refinement steps) needed to increase quality and coverage of the collected knowledge repository.