Meaningful Texts: The Extraction Of Semantic Information by Geoff Barnbrook, Pernilla Danielsson, Michaela Mahlberg

Show description

To process a hypertext in a corpus we have to know where it begins and 26 M E A N I N G F U L TEXTS where it ends, hence the dynamic property has to be handled and the imported file can only represent a static picture of the hypertext and pay attention to modularization and linkage. Even if this approach does not allow us to integrate all aspects of hypertextual features, the mediumspecific dimensions can be captured to a certain extent. A third major problem subsists with regard to the inclusion of interconnected visual information into the corpus, particularly strongly blended forms of inter-semiotic layering.

G. g. g. the CORIS/CODIS project for written Italian or the corpus of spoken Israeli Hebrew). In any case, the non-English language communities, even if they are prestigious, still lag behind the advances of corpus linguistics with reference to English. g. general questions to be A D I A C H R O N I C GENRE CORPUS 21 answered in contrastive syntax, morphology or textology). For now there is little possibility of comparing results across corpora, neither within the same nor between different languages and certainly not in terms of language change.

Tagging relies on what one might call the dictionary model of word representation, namely, that word senses are discrete and enumerable. The dictionary model predicts that annotation is easy: taggers inspect the occurrences of a (polysemous) string in a corpus, interpret and determine its meanings, and match these against a dictionary entry. Tagging should be easy, since it should mimic our everyday behaviour of processing language input and looking up entries, as it were, in our mental lexicons.