title = {Deep and Shallow Processing of Modern {Greek} in a Multilingual Context},
 author =  {Julia Neu},
 school = {Saarland University, Computational Linguistics Department},
 year = 2004,
 address = {Saarbr{\"u}cken, Germany},
 abstract = {
Systems for deep and shallow processing have proved widely successful for the syntactic and semantic analysis of different European and Asian languages. In that respect, it has turned out that the deep and shallow techniques show complementary strengths and weaknesses. Until now, however, there has only been done little practical work on Modern Greek, especially with respect to the development of grammar resources.\\
    This thesis provides an account for deep and shallow processing of Modern Greek in a multilingual context. First, I describe the development of a reusable, multi-purpose linguistic resource for Modern Greek: a deep computational Modern Greek HPSG grammar. The deep processing results of a set of Modern Greek test sentences are illustrated and evaluated in detail.\\
    Second, the development of the SProUT resources for purposes of shallow processing, in this case the components for tokenization and Named Entity Recognition in Modern Greek, are presented and exemplified afterwards by analyzing texts about special domains taken from the internet.\\
    The third and last part of the thesis deals with the combination of both the deep and the shallow grammar resources. It will be illustrated how Robust Minimal Recursion Semantics (RMRS) can work as an interface between the two levels of deep and shallow analysis. Moreover, by giving concrete examples, I will demonstrate how shallow processing with SProUT can fill in the gaps which may arise from deep processing with the Modern Greek HPSG grammar.