This page describes my PhD research which investigates the use of inheritance networks to build multilingual lexicons. Research into the application of inheritance networks to multilingual lexical description is relatively new and many theoretical and methodological issues are still unanswered. My thesis focuses on three of those issues:
- The regulation of the inter- and intralanguage inheritance relations
In a multilingual inheritance network, there are not just inheritance relations within a language, but there are also inheritance relations between languages. The question is how are these two kinds of inheritance relations expressed and how do they interact? In other words, how does the hierarchical structure of one language interact with the hierarchical structure between the languages?
- Multilingual information sharing
This issue can be subdivided into three questions. First, which information can be shared in a multilingual inheritance lexicon and can different architectures share the same information? In theory, a multilingual inheritance network can share information at all levels of linguistic description -- syntax, semantics, morphology, phonology, etc. Second, which information should be shared in a multilingual inheritance hierarchy? Languages exhibit similarities for several reasons -- genetics, typology, language contact or chance. Should a multilingual lexicon capture all these similarities or should only a specific kind of similarities be captured? Third, how is the information shared in a multilingual inheritance network? That is, which criteria are used to decide when information should be stated in a shared part?
- Development strategies
How does one go about constructing a multilingual inheritance lexicon? Should the monolingual and multilingual hierarchical lexicons be developed in parallel and linked immediately upon construction or should a non-parallel development strategy be adopted, where the monolingual lexicons are first fully developed separately and only linked together at the end?
These issues are explored by comparing different architectures for multilingual inheritance lexicons following Evans' proposals. Evans distinguishes what he calls parameterised and non-parameterised architectures for multilingual inheritance-based lexicons. In a parameterised model, linguistic descriptions are made multilingual by conditionalising the linguistic objects for language. The information that is valid for one particular language is then the collection of specifications that mention that language. In a non-parameterised model, on the other hand, language is not explicitly used as a parameter. The language-specific parts are separate hierarchies which are linked together through a common hierarchy which contains what the language-specific hierarchies have in common. There is nothing in this model that ties linguistic objects explicitly to a particular language. You have to know which hierarchy you are in to know which language you are talking about.
In order to compare the different architectures and to find out what their advantages and disadvantages are, sample lexical fragments have been implemented. The lexical fragments are implemented in DATR, an inheritance-based formalism developed by Evans and Gazdar (1996), and focus on a subset of the Germanic languages including Dutch, Danish, English and Icelandic. A sample fragment can be found here.
Work that is closest in spirit to my research is the PolyLex project (Cahill and Gazdar, 1995 - 1999) which produced a hierarchical multilingual lexicon for Dutch, English and German. The PolyLex project focused on the sharing of morphological, phonological, and morphophonological similarities. More information about PolyLex can be found on the PolyLex webpages.
This page is maintained by Carole Tiberius and was last modified in October 2001.
| Home Page | | Links |