Preparation and Usage of Xhosa Lexicographical Data for a Multilingual, Federated Environment

Preparation and Usage of Xhosa Lexicographical Data for a Multilingual, Federated Environment Sonja Bosch, Thomas Eckart, Bettina Klimek, Dirk Goldhahn, Uwe Quasthoff Department of African Languages, University of South Africa; Natural Language Processing Group, Agile Knowledge Engineering and Semantic Web Group, Leipzig University boschse@unisa.ac.za, {teckart, klimek, dgoldhahn, quasthoff}@informatik.uni-leipzig.de Abstract The South African linguistic landscape is characterised by multilingualism and the influence between their eleven official and some local languages. Unfortunately, for most of the languages the amount and quality of available lexicographical data is suboptimal, even though its availability is essential for all educational institutions and for the development of state-of-the-art language technology. In this paper we present a new source of lexicographical data for Xhosa, a language spoken by more than eight million speakers. For its utilisation in a multilingual and federated environment it is modelled using a dedicated OWL ontology for Bantu languages and possesses all features that are currently considered integral for the promotion of resource reuse as well as long-term usage. In the future, the introduced on- tology may be used for other Bantu languages as well and may ease their combination to achieve more extensive, multilingual data stocks. Keywords: Xhosa, lexicography, research infrastructures, linked data 1. Introduction A basic requirement for the language processing capabil- ity for any language is the availability of lexicographical data, ideally open source data which is often hard to find for less resourced languages. This includes many members of the Bantu language family. For enhancing the usabil- ity of this kind of data this paper presents a Bantu Lan- guage Model for describing lexicographical data in RDF. Furthermore, it presents a new resource of lexicographical data for the Xhosa language based on this new model. The data to be presented in this model is a representative sam- ple of raw data for a Xhosa-English dictionary, containing approximately 6,800 lexical entries. In its final state, the data set should contain approximately 10,000 lexical en- tries. Whereas the available data enables us to use Xhosa as the language of instantiation, the method and model are extensible and applicable to many other Bantu languages, in particular those belonging to the same group. Together with this paper, both the Bantu Language Model and the current state of the lexicographical data set are freely available for download and for querying via a ded- icated SPARQL endpoint. It should be noted that the re- search reported on in this paper is work in progress. Its final version will be provided via SADiLaR, the South African Centre for Digital Language Resources. Moreover, the cur- rent version of the data is already available via CLARIN-D (see section 4.). The remainder of this paper is structured as follows: Sec- tion 2 describes the origin of the Xhosa language material and explains essential features of the Xhosa language with a focus on its morphology. Section 3 explains the new Bantu Language Model that is based on the established MMoOn ontology1. Section 4 gives detailed information about the structure of the …

Content truncated to 500 words