Inducing Knowledge Graphs using Zero Shot Slot Filling
A main barrier for the adoption of Knowledge Graphs (KG) for enterprise is the large effort required to build and populate them. In this talk, we focus on the slot filling task: given an entity query in form of [Entity, Slot, ?], a system is asked to 'fill' the slot by generating or extracting the missing value from relevant passages. This capability is crucial to create systems for automatic knowledge base population, which is becoming in ever-increasing demand, especially in enterprise applications. We present the results of our long standing research direction in this area. Specifically, we will present our recent advances in applying end-to-end retrieval-based language models to zero-shot slot filling tasks, achieving state of the art performances in public leaderboards.
Alfio Gliozzo is a researcher with over 20 years of experience in the field of Artificial Intelligence. He manages the Knowledge Induction department at IBM T.J. Watson Research and he is a principal researcher and global leader for the area of Knowledge in IBM research AI. His research focuses on automatic induction of knowledge graphs from text and their exploitation for enterprise search solutions. He published approx 100 scientific publications, including books, papers and patents. He was part of the Deep QA team that developed Watson, the system that defeated the Jeopardy! grand masters.