next up previous contents
Next: Standardising Subcategorisation Up: Syntactically annotated corpora Previous: Phrase and clause level

Preliminary Recommendations

Summary

In table 3.11, the various corpus syntactic annotation schemes are summarised with respect to the type of linguistic information which may be included in the lexical representation of verb entries, and their subcategorisation frame. The table is provided to clearly correlate the markings in an annotated corpus with the information that may be provided in a lexicon, and in doing so may seem to blur the distinction between types of information that are relevant to either a lexicon, or an annotation scheme, which in most cases will be very different. However, the feeding of information between the lexicon and the annotated corpus can be seen as bidirectional -- much of the information marked in a corpus will necessarily be included in the corresponding lexicon, but this information is then derivable from the corpus, for enrichment of the same, or another lexicon.

 

ENGCG IBM TOSCA SUSANNE UPENN
arg-no + infer infer infer infer
syn-cat infer + + + +
funct-role + - + + +
control tex2html_wrap8271 - + + +
lex-selecttex2html_wrap8273 + + + + +
deep-str - - - + +
morphsyn-constr By the nature of an annotated corpus, this information will be inferable
frame-alt By the nature of an annotated corpus, this information will be inferable
Table 3.11: Subcategorisation information in corpus annotations 

tex2html_wrap8271Although grammatical function information is included, control information is not necessarily deriveable since ambiguity may remain in the final analysis.
tex2html_wrap8273In an annotated corpus such phenomena will be marked either as a (bound) particle, or if not discontinous, then possibly as a multi-word verb form.



next up previous contents
Next: Standardising Subcategorisation Up: Syntactically annotated corpora Previous: Phrase and clause level