... Tübingen^2.1

©2003, Detmar Meurers, Thilo Götz and Dale Gerdemann.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... editor.^2.2

Also, be aware when mixing tabs and spaces: If you type in a space and then a tab, the space may become invisible on the screen, but still counts when calculating the indentation level!

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... consistently.^2.3

We feel that the advantage of this approach is that it allows the hierarchical nature of the type hierarchy to be seen without too much obscuring syntax. A purely graphical type hierarchy interface would be even better, but this is as close as we could come to a graphical presentation within the confines of ASCII. It must also be admitted that using level of indentation to indicate subtypes a certain level of complexity. It is no longer possible to give a BNF definition of the type hierarchy since the level-of-indentation must be passed around as a context sensitive feature.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... (&).^2.4

The ampersand is not necessary but it is recommended. Our experience has shown that unintended multiple inheritance, by accidently using the same type name for different purposes, is a common error. If the ampersand is omitted, the system will give a non-fatal error message.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... matrices,^3.1

Note that the corresponding Boolean algebra for these matrices is a semiring where $1\oplus 1=1$ because as mentioned above, in a Boolean semiring, the additive operator $\oplus$ corresponds to OR.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...,^3.2

$\mathord{\downarrow}$ means defined.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... University^8.1

©2003, Vanessa Metcalf, Detmar Meurers and Markus Dickinson

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... each.^8.2

We actually translate a copy of each description into a feature structure, since desc2fs/3 adds type information to Prolog variables, thereby making them unrecognizable as such to var/1. We rely on Prolog variables in the descriptions to signal the presence of a path equality in the lexical rule descriptions. We keep track of this information in order to avoid creating undesirable path equalities in the frames.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... succeeded.^8.3

To obtain the correct follows relation for a lexical rule, however, information from the input specification should to be transferred to the output, ideally via a generalized frame predicate. Currently, we include a lexical rule on the follows list if unification succeeds utilizing at least one of the clauses defining the frame predicate for the lexical rule in question.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... chart:^9.1

Note that for ease of exposition, we are showing a simplified version of the code here.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... below:^10.1

VF stands for Vorfeld, CF for Complementiser Field (left sentence-bracket), MF for Mittelfeld, VC for Verbal Complex (right sentence-bracket), and NF for Nachfeld

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... Prolog.^12.1

The above information has been adapted from http://wombat.doc.ic.ac.uk.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.