UW CSE AI Seminar 2016 - Ashish Sabharwal, Beyond Informational Retrieval
Explore the challenges and advancements in artificial intelligence for answering science questions in this 59-minute seminar by Ashish Sabharwal from AI2. Delve into the complexities of combining multiple facts, utilizing common-sense knowledge, and employing semi-structured reasoning techniques to tackle elementary-school level science exams. Learn about a novel approach using Integer Linear Programming (ILP) that surpasses traditional information retrieval methods and probabilistic logic systems in robustness, scalability, and accuracy. Discover the fundamental challenges of science question answering, progress made in the field, and future obstacles to overcome. The seminar covers topics such as ensemble solvers, semantic constraints, automatic explanation generation, knowledge graphs, and diagram questions, providing a comprehensive overview of the current state and future directions of AI in science education.