A Synergistic Approach to Extraction, Learning and Reasoning for Machine Reading.

Artificial intelligence researchers have long been attracted by the possibility of designing systems that can learn by reading textual information. In order to meet this goal a system must be able to (1) extract information from text, (2) learn rules that fuse together different pieces of information to make novel inferences and (3) reason about which extracted and inferred facts are true. To date, researchers have made significant algorithmic advances within each of these disciplines. However, there has been far less work that addresses all three problems with a single framework. The goal of this project is to explore a more unified machine reading system, with the overall aim of increasing the systems performance. We plan to pursue the following two key scientific objectives. The first objective is to develop a method that allow for a multidirectional synergy between the algorithms used for extraction, learning and reasoning in a machine reading system which permits rich feedback between the different modules. The second objective is to pursue a sophisticated and rigorous approach to managing the inherent uncertainties in extraction, learning and reasoning in a uniform manner.

Keywords:Algorithms, Machine learning

Disciplines:Applied mathematics in specific fields