Horizontal, vertical and other formats (2024)

Horizontal, vertical and other formats

We must, for practical as well as definitional reasons,restrict our attention to corpora considered as collections oftexts or textual samples of language. Texts are linear;syntactic structures, on the other hand, are often representedin two-dimensional terms, especially as tree structures, or(in greater detail) as tree structures, the nodes of which aresets of attributes and values. As far as syntactic annotationis concerned, we are interested only in how these two- ormulti-dimensioned structures are represented in relation tothe linearity of texts.

There are two general commonly-used linear formats for storing,inputting and outputting text data: horizontal and vertical. It ispossible to represent a syntactically annotated text in either ofthese formats, without changing the nature of the annotation. Theconversion of a horizontal to a vertical format or vice versa is arelatively trivial operation if undertaken automatically. However, fromthe user's point of view, the difference between the two formats iscertainly not trivial, as it may make the difference between anintelligible and an unintelligible presentation. We will use examplesfrom some corpora to illustrate this.

The first example is from the Associated Press Corpus withLancaster skeleton parsing annotation. The sentence in 2 can be representedin a horizontal format, as in table 1.

The labelled bracketed analysis can be represented in avertical format, as in table 2.The original sentence is in the first column, the part-of-speech tags inthe second, and the brackets and labels constituting thesyntactic annotation appear in the third column.

Table 3 is an example in horizontal format from the IBM Paris Treebank (Langé 1994).

The horizontal format is more compact, and is easier to read so longas the amount of syntactic information interspersed with the words isnot too dense. The vertical format is more convenient and morereadable if there is too much syntactic information to beconveniently shown in the horizontal format. Moreover, the verticalformat lends itself to a number of parallel fields of information, sothat (for example) the actual orthographic text (as a sequence ofword forms and punctuation marks) can be separated out from thesequence of morphosyntactic tags, and both of these separated fromthe representation of a phrase structure tree. Other fields maycontain corpus location references, and deep syntactic information(such as ellipsis) alongside in a separate field from the surfacesyntactic information. Table 4 is an example from the SUSANNEcorpus (Sampson 1995), which gives an impression of the variousaligned information types that can be given. The columns (i.e fields)contain the following information:

Field 1:

text references

Field 2:

part-of-speech tags

Field 3:

the text words

Field 4:

base-form (lemmatised forms of Field 3; e.g. said is lemmatised as `say')

Field 5:

syntactic annotation (brackets and labels)

The field that indicates the structure of the sentence can bemade more graphically explicit by the use of indentation. Theexample from TOSCA in table 5illustrates this. On the first level isUtterance, the second level NP, VP and PP, and so on. (Thisindented format is in fact an intermediate structure, the final outputbeing represented as a tree on the screen.)