Symbolic Learning and Reasoning

Dr. Nil Geisweiller is leading a team carrying out advanced R&D on symbolic learning and reasoning in the IBM Watson framework. The high-level motivation and conceptual background of this work is covered in research blog posts such as Introspective Reasoning Within the IBM Watson Framework and Enabling Cognitive Visual Question Answering.

This work involves integrating multiple AI tools, such as the probabilistic logic networks (PLN) logic engine, the MOSES automated program learning engine, the IBM Watson pattern miner, and the ECAN attention allocation system, into a common framework based on IBM Watson unified rule engine (URE).

Conceptually, the key theme is leveraging reflective meta-learning and cognitive synergy (win–win interoperation between different cognitive algorithms) to achieve higher levels of generalization and abstraction in machine learning/reasoning.