Registration is open from 09:45 (grab some coffee!).
The tutorial starts at 10:15 in room T03 R02 D81.
Information on the hands-on parts:
Lexical Resources for Natural Language Processing
The goal of this tutorial is to gain a fundamental understanding of lexical resources, their structure, information types, and potential for a wide range of natural language processing applications.
Lexical resources play an important role ever since our scientific field came into being. They are used as multilingual translation dictionaries in machine translation, as a source of background knowledge for parsers and morphological analyzers, as a knowledge base for named entity detection, logical reasoning, and question answering, as a sense inventory for word sense disambiguation and lexical substitution, and they are almost indispensable for natural language generation.
Working with lexical resources is, however, often challenging. There are many different resources available, and each of them has a different, often highly complex structure requiring training time and careful adaptation of the software tools that should benefit from their knowledge. Apart from that, lexical resources are limited in their coverage and the type of lexical knowledge they encode.
In this tutorial, we address these challenges in two major learning units. We first explain the theoretical background by introducing a broad range of different lexical resources including WordNet, GermaNet, FrameNet, VerbNet, OpenThesaurus, OmegaWiki, Wikipedia, Wiktionary, and OntoWiktionary. To overcome the problems of different structure and coverage, we additionally discuss state-of-the-art solutions of inter-linking and standardizing lexical resources, and we put a special focus on the type of information found in a specific source.
In the second learning unit, we explore the contents of different lexical resources in multiple practical exercises. We use the example of the linked lexical resource UBY (http://www.ukp.tu-darmstadt.de/uby), which provides lexical knowledge from all aforementioned resources using a unified data model. We explain the underlying principles of the Lexical Markup Framework (LMF), which has been standardized as ISO 24613 and acts as a meta model for UBY. Obtaining and accessing different kinds of lexical knowledge using the UBY API is the main goal of this learning unit, which we achieve by developing and running multiple code examples. We particularly invite our audience – upon interest – to try the code live on their personal notebooks.
A guiding motive in both the first and the second unit are different application scenarios, such that the knowledge gained in this tutorial can directly be applied to one’s own research. Besides well-established knowledge-based approaches, we also take a glance at emerging trends of using lexical knowledge bases in weakly supervised and hybrid methods.
This tutorial is suitable for master- and PhD-level researchers who have no or little experience with lexical resources as well as for seniors, who plan to complement their research projects with lexical knowledge or who are seeking for innovative ways of teaching lexical resources using a “hands-on” approach.
Christian M. Meyer and Hatem Mousselly Sergieh
Ubiquitous Knowledge Processing Lab
Technische Universität Darmstadt