Arrernte Footy: Natural Language Generation for AFL Reports in Aboriginal Languages

Mark Dras, Department of Computing, Macquarie University will speak on "Arrernte Footy: A Computational Grammar and Natural Language Generation System for Arrernte" at the Australian National University School of Language Studies in Canberra, 4pm, today. He will discuss the analysis of reports of AFL football matches in the language of the Arrernte people of Mparntwe (Alice Springs).

ANU School of Language Studies
Monday 13 February SEMINAR 4 pm Baldessin Building, room W 3.03 (Map)

Mark Dras, Department of Computing, Macquarie University

Title: A Computational Grammar and Natural Language Generation System for Arrernte

Abstract:

Natural Language Generation (NLG) systems have as their goal the generation of human-like text from some underlying representation, often numerical or other data from a database. We are currently developing an NLG system to generate texts in both English and Arrernte about Australian Rules football games, of the sort found in newspapers after a game has been played (e.g. "GEELONG has claimed the club's ninth premiership and marked itself as one of the greatest sides of all time with a stunning 38-point Grand Final win against Collingwood...."). The final stage of this system is a realisation component that is built around Lexical Functional Grammar (LFG) and its XLE environment for grammar development: this takes a description of the functional structure of a sentence, and uses a computational grammar to make appropriate lexical and syntactic choices to produce the actual sentence. A large-scale grammar of English has already been developed by the international ParGram project, which we will be using; we are now building a computational grammar of Arrernte, based on existing linguistic descriptions of the language.

In the talk I will give an overview of this system, with a focus on the computational grammar of Arrernte, in particular how we've implemented in the LFG formalism a number of constructions that have been discussed in the literature, such as complex predicates like associated motion.