Period: 1998-2000
The goal of this research is to study an integrated theory of learning, knowledge representation and reasoning and evaluate it on large scale knowledge intensive inferences in the natural language domain. Recent studies, within the Learning to Reason framework, have shown that there is much to gain from studying these issues within a unified framework. This research investigates some of the fundamental issues within this framework -- concentrating on a probabilistic setting. The emphasis is on developing algorithms that exploit the relaxation of some of the "traditional" assumptions in this domain. These include requirements put on the learning algorithms (e.g., learn a "good" density estimation), reasoning algorithms (e.g., support *all* queries uniformly well) and on some of the knowledge representations studied in this domain. Application of these theories to the natural language domain are studied and evaluated experimentally. The emphasis is on learning methods and representations for combining low-level learning algorithms to perform higher level inferences. This research will have impact both on understanding some of the fundamental issues involved in combining learning and reasoning and will allow for making concrete progress towards bridging the gap between the low-level work and higher level goals in the natural language domain.