From NLP to NLU: Why We Need Varied, Comprehensive, and Stratified Knowledge - Neuro-Symbolic AI

Explore the critical role of knowledge in advancing natural language understanding (NLU) and addressing limitations of current language models in this 55-minute seminar presented by Prof. Amit Sheth at USC Information Sciences Institute. Delve into the importance of varied, comprehensive, and stratified knowledge in neuro-symbolic AI systems to overcome challenges such as hallucinations, lack of recency, and limited explainability. Examine the multifaceted nature of knowledge required for deeper language understanding, including lexical, linguistic, common sense, and domain-specific knowledge. Learn about different approaches to knowledge infusion and elicitation in AI systems, and discover how combining neural networks with symbolic components can lead to more robust and interpretable AI solutions. Gain insights into the development of knowledge graphs and their potential to enhance AI capabilities in areas like semantic search, clinical NLP, and emotion analysis.

Welcome to the Al Seminar Series
Challenges with Current LMs
Challenges to be addressed for NLU
Knowledge (Graphs) to the rescue
Characteristics of Knowledge Infusion
Knowledge Verified Interpretable Prediction Al through Process Knowledge Structures
Knowledge Verified Interpretable and Safe Text AIS Generation through Process Knowledge Structures
Takeaway
Knowledge Graph Development Challenges & tool functionalities we are working on