next up previous contents
Next: Guidelines Up: Layers of annotation Previous: Information about spoken language

Recommendations

Conclusion

Layers (a)-(h) represent a loose hierarchy of informational details, and in some cases a `higher' type of information cannot be added unless a `lower' (i.e. more basic) type has already been provided: e.g. (a) is a pre-requisite for (b).

Table 9 below summarises which of the features distinguished in (a)-(h) have been implemented in a number of annotation schemes. (It is a `quick overview'; a more detailed overview is given in the documents SASG2 and SASG3.)

 

tex2html_wrap3982 tex2html_wrap3984 tex2html_wrap3986 tex2html_wrap3988 tex2html_wrap3990 tex2html_wrap3992 tex2html_wrap3994 tex2html_wrap3996 tex2html_wrap3998
Bracketing (a) x x x x x - x x x
Labelling (b) x x x x x - x x x
Dependency relations (c) - x* x* - - x x - -
Syntactic functions (d) - x x (x) x x** x x x*
Subclassification (e) - x - - x x* - - x
Logical information (f) - x x - - x* x - -
Rank (g) x x x - x - x x -
Spoken language (h) - x* - - x* - x x x*
* intended for future release
** at word level
- not implemented
Table 9: Overview of annotation levels in selected corpora 

However, there is no strict precedence between the levels, and many combinations of information types, other than those represented in the table, are in principle possible. It is conceivable, for example, that someone might wish to indicate functional information about syntactic segments in a corpus while ignoring other kinds of information.



next up previous contents
Next: Guidelines Up: Layers of annotation Previous: Information about spoken language