Denoising automatically acquired knowledge graphs for domain specific knowledge-based applications
Date: Tue 9 Nov 2021
Speaker: Stefano Faralli
Title: Denoising automatically acquired knowledge graphs for domain specific knowledge-based applications
Abstract: In recent years, research in information extraction and knowledge acquisition produced knowledge resources on a scale that was arguably hard to imagine just a few years ago.
Web-scale open information extraction systems have been successful in acquiring massive amounts of machine-readable knowledge by effectively tapping large amounts of text. Moreover, a variety of methods have been developed to automatically acquire fully-fledged knowledge bases of ever-increasing coverage and complexity directly from the Web. The availability of vast amounts of knowledge encoded in heterogeneous sources, including both structured, semi-structured as well as unstructured resources, has paved the way to methods for the induction and/or integration of a wide range of different taxonomic resources, ranging from domain-specific taxonomies all the way through collaboratively created ones.
All in all, the resulting automatically acquired knowledge can achieve good coverage and domain-specificity and consequently enable high-end applications.
In this talk, I will present some recent research directions that I have been exploring in the field of taxonomy induction from automatically acquired noisy knowledge graphs.