Coding Units Sample Clauses

Coding Units. The coding units are the atomic units which are used to define the smallest element of the annotation efforts. This element should be defined by both using the distinction criteria, which are defined by Xxxxxxxxxxxx (2004a, Chapter 5), and the theory behind the annotation effort. In the TDB, the theory of annotation does not sanction any coding unit directly, because it is a low-level theory which does not limit the annotators by unit or structure definitions. Therefore, the criteria, which are used while determining the coding unit, should also not commit themselves to any high-level proposal. Therefore, the coding unit of the TDB should be the smallest piece of text. When all these are considered, there are two strong candidates; one of them is selecting characters as coding units as a result of a physical distinction criteria evaluation, and the other is selecting words as coding unit as a result of syntactical distinction. As mentioned before, the TDB annotations are performed by using a computer program which is called as the XXXX (Aktaş et al., 2010). In the XXXX, the annotations are performed by selecting text spans character by character. Also, the annotated text spans are represented by character offsets relative to the starting character of the text. The offsets of starting and ending characters are held for each discourse element (Arg1, Arg2, and Conn) annotation. That is to say those, characters are physical distinctions that are automatically generated by the TDB annotation effort. When it is considered from the statistical perspective, a character is an atomic unit for texts that cannot be split up any into smaller pieces, which is an important property that is necessary to sustain the statistical representation capability of the unit. However, one would assume that in the TDB annotations, annotators do not decide to include or exclude a character into a discourse element, but they decide if they should annotate a word or not. An annotation, which a part of word is included into the annotation and the other part is not, is mis-annotation except the subordinator connective annotations. A part or whole of the subordinator can be morphologically embedded into the last word of the Arg2, i.e., the verb. Yet, this is not a serious obstacle for taking words as coding units as this situation can be handled with a preprocess phase prior to agreement measurement. For the statistical perspective using words as coding units are also plausible since a wor...