Ontologies and ontological methods in linguistics

Lecturer: 
Topic: 
Language & Computation
Level: 
Introductory
Abstract: 

An ontology is a cross-connected network of relevant concepts that makes explicit, classifies, and organises the assumptions and terms of the domain in question. The notion of ontology has become rather prominent in linguistics over the last decade, and, through the development of Semantic Web technologies, it has received increased attention in artificial intelligence and computer science more generally. This course looks at ontologies in and for linguistics from both a theoretical and an applied perspective. In theoretical linguistics, an ontological approach allows (i) to formally model the speakers’ cross-connected network of linguistically relevant concepts (‘speaker ontology’) – and thus enables the modeling of semantic and grammatical relationships in a fine-grained way, and (ii) to make explicit the linguistic scholars’ conceptual network – i.e. to model the domain-specific ontology of the general linguistic description apparatus (‘discipline ontology’). From a computer science perspective, Semantic Web technologies allow us to integrate language resources and make them collectively accessible and searchable (resulting in linked data in linguistics). To understand the notion of ontology is thus to not only understand the foundation on which the Semantic Web is built. It is also to understand how discipline notions are cross-related to one another (with linguistics being the discipline under discussion) and how objects of study may give leads to an underlying ontology, i.e. how language data showcase underlying concepts entertained by speakers of these languages. This course addresses all of these aspects. It introduces the notion of ontology and practices the tool-supported building of an ontology. Speaker ontologies as well as discipline ontologies are introduced and discussed. For purposes of exemplification, different languages (to compare speaker ontologies) and different linguistic theories (to compare discipline ontologies) are deployed. Furthermore, two case studies illustrate the usage of ontologies: Case study (1) discusses the Typology Tool (TYTO), a resource under development that aims at straddling across both speaker and discipline ontologies. Focusing on the domain of social cognition, TYTO endeavours to systematically capture the different concepts surfacing cross-linguistically in the domain. The underlying ontological structure of TYTO is introduced and arising issues are discussed. Case study (2) addresses the idea of making language resources semantically interoperable. A prime example of this is the Australian National Corpus (AusNC), which aggregates collections of language data from a diverse range of sources. While search and retrieval functions are now technically available across the entire aggregated data set, annotations of the different collections vastly differ, making such a unified search capability problematic. One solution to such a problem is to construct an interlingua ontology which explicitly represents the knowledge embodied in the annotations of all the collected data. It is exemplarily shown how this can be done and which improved functionality the AusNC would gain if a comprehensive annotation ontology existed.

Week: 
First week
Slot: 
17:00 - 18:30 - slot 4