View Item 
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        JavaScript is disabled for your browser. Some features of this site may not work without it.

        Browse

        All of UU Student Theses RepositoryBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

        How to choose article and number for nouns when translating from Japanese to Dutch in the context of syntax-based machine translation

        Thumbnail
        View/Open
        versie2thesis.pdf (1.772Mb)
        Publication date
        2014
        Author
        Broers, L.D.S.P.
        Metadata
        Show full item record
        Summary
        When translating from a language which does not have distinction in number (singular/plural) or an indication of definiteness for nouns (Japanese), into a language which does have these distinctions (Dutch), it is not straightforward to make these decisions. This problem in machine translation is studied using a compositional translation framework. This framework uses synchronized (isomorphic) grammars to deal with translation. As the grammars are reversible, no separate analysis and production for both languages are necessary. The basic expressions and rules in both languages correspond in meaning, and when rules with the same meaning are applied in both languages, the resulting sentences are each other's translations. The grammar also uses features to indicate syntactic and semantic properties of the words and phrases. The rules check for certain settings of the features to decide whether or not they are applicable. If necessary, features of the resulting output expressions are set as well. A rule also has conditions, which all have to be met before a rule can be applied. These restrictions are intended to prevent ungrammatical combinations from being generated. The features `divisible', `generic' and `count/mass noun' are the ones the rules make the most use of in deciding number and definiteness for nouns. The output created by the rules consists of all possible translations. A bonus system is used to order the possible translations by plausibility.
        URI
        https://studenttheses.uu.nl/handle/20.500.12932/19145
        Collections
        • Theses
        Utrecht university logo