Natural language Processing (NLP)

Marieke van Erp

It is estimated that about 80% of knowledge is captured in human (or natural) language, such as news, social media, wikis and books. To facilitate access to the information contained in text, computers need some capabilities to ‘understand’ human language, but this is not as easy as it seems.
In this lecture, I will cover the phenomena that make language difficult to computers as well as give an overview of the core natural language processing techniques and tools developed in the field of computational linguistics. The focus here will be on traditional natural language processing pipelines, with a short sidestep to deep learning methods for NLP. But the first lesson is: NLP ≠ deep learning.

Ontology Design Patterns and Linguistic Frames

Aldo Gangemi

Human knowledge is extremely nuanced, contextually-dependent, and takes many forms. Yet, some regularities do exist, otherwise we would not be able to communicate and collaborate: the physical, biological, social, and cognitive worlds in which we live are the limits that we experience during our lives, and eventually make our knowledge emerge. We can study regularities with reference to those worlds, using computational means: What are the semantic theories that can represent regularities? How to use them in semantic applications? How to measure their validity, commonality, interoperability? Patterns, frames, and reasoning workflows will be presented, as they are empirically designed, discovered, composed, and as they contribute to qualitative applications.

Human in the Loop

Irene Celino

The advent and rise of machine learning technologies, as advertised on media, often leave the audience with the impression that data processing and predictions are completely automatic and driven by autonomous agents. The reality that every data scientist knows and experiences is quite different: human intervention is still needed and often indispensable in several steps of the research process, from collection to labelling, from processing to validation.
The Human-in-the-loop “pill” will focus on giving an overview of methodologies, techniques and applications through relevant examples: the talk will distinguish the approaches (crowdsourcing, citizen science, human computation) as well as the different objectives (model validation, prediction explainability, algorithmic transparency).

Event Timeslots (3)

Tuesday 2nd
-
Natural Language Processing

Wednesday 3rd
-
Ontology Design Patterns

Wednesday 3rd
-
Human in the Loop

Categories: