Talk

I am giving a talk at the Informatics Doctoral Research Forum, Indiana University.

Date and Time: 2 of Nov. 2019, 10:30 AM

Location: Wells Library, Hazelbaker Hall

On Knowledge Representations and Reasoning for AI using Deep NLP and Description Logic

We will discuss examples of mapping content to knowledge representations and Knowledge Graphs using broad and deep Natural Language Processing (NLP). The identification of entities, relations between entities, and generation of graph-based representations is one of the problems we address. Linking entities to knowledge graphs (e.g. YAGO, DBpedia, ConceptNet) and various other forms of knowledge representations (e.g. Wolfram Language 12 entities) is the second problem we will discuss.

To achieve acceptable entity/relation extraction and entity linking, we rely on NLP that is often shallow, with varying margins of error, unreliable with complex constructions and texts. We will address ways to compensate for parser errors, anaphora resolution mistakes, or entity labeling issues and discuss hybrid methods that integrate data and knowledge driven deep and reliable NLP components in a complex system for text 2 knowledge graph mapping.