Integrating Classical Sindarin Morphology and Syntax using XFST and XLE

Sindarin is a constructed language invented by the author and linguist J.R.R. Tolkien. It is the language spoken by the Elves. Though the number of complete sindarin sentences is rather limited[1] enough information is available to derive syntactic rules and build a Sindarin grammar. This project aims at modelling the basics of Classical Sindarin grammar in LFG using Xerox XLE system. In addition it interfaces to Xerox XFST to lookup morphological information which has already been implemented in another project. For more information about Classical Sindarin morphology see the project documentation[2]. The project is mainly based on information provided by David Salo’s ”A Gateway to Sindarin”[1]. 1 System Overview Figure 1 shows an overview of the system implementation. It consists of three components which are further discussed in more detail. 1.1 Tokenizer The tokenizer takes textual input and generates tokens that can be further processed and analysed by the morphological analyser. It performs the following tasks: – all characters are cast to lower-case. – accents are removed. – binding articles (like i-) are split from nouns. – numbers are identified and isolated – special and punctuation characters characters are isolated. – identified tokens are separated by a token boundary TB. Tokenization removes ambiguities and simplifies the creation and maintenance of lexicon files for the morphological analyser and the xle grammar. Though xle provides a basic tokenizer that is suitable for english and german grammars it is not sufficient for parsing Sindarin. The main reason is the existence of articles that bind directly to the corresponding noun. They have to be split and 2 Thomas Zink, Andreas Engl, and Wolfgang Miller