...Gazdar

Cognitive & Computing Sciences, University of Sussex, Brighton BN1 9QH, U.K.; Lynne.Cahill/Gerald.Gazdar@cogs.sussex.ac.uk

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

...here

See Fabri & Wunderlich (1995) on verbs; Bleiching (1992) on nouns; and Bleiching (1994) on nouns and verbs. Wurzel (1989, 86-90) does discuss article inflection, but only as a diagnostic for noun inflection.

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

...descriptions

See Ejerhed (1995) for a proposal.

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

...central

As in the work of Matthews (1972), van Marle (1985) and Carstairs (1987).

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

...Morphology

See Brown et al. (1996), Brown & Hippisley (1994), Corbett & Fraser (1993), and Fraser & Corbett (1995; in press) for recent work in this framework.

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

...Morphology

See Stump (1992; 1993a; 1993b; 1993c; 1995) for representative work in this framework, and Gazdar (1992) for discussion of formalization issues.

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

...Bielefeld

See Bleiching (1992; 1994), Gibbon (1990; 1992), Gibbon & Bleiching (1991), Langer & Gibbon (1992) Reinhard (1990) and Reinhard & Gibbon (1991) for examples of this work.

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

...forms

Bleiching (1992) goes some way towards this idea, specifying lexical stress placement in compounds by referring to the phonological and morphological structure. Gibbon (1992, 48) takes the same step in his treatment of umlaut. And Ulrich Wille's unpublished 1992 work on German verb inflection follows a similar path.

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

...implemented

There have been more than a dozen different implementations of the DATR\ language. They include Evans's (Brighton) implementation, which is written in Prolog and runs on most Unix platforms; Gibbon's (Bielefeld) DDATR Scheme, NODE Sicstus Prolog, and awk implementations; Kilbury's (Düsseldorf) QDATR Arity, Quintus and Sicstus Prolog implementations; and Illouz's (Paris) implementation of CDATR (in C). All of these are freely available on request, as is an extensive archive of over one hundred and eighty example fragments some of which illustrate formal techniques and others of which are applications of DATR to the lexical phonology, morphology, syntax or semantics of a wide variety of different languages (including nontrivial fragments of aspects of the lexicons of Arabic, Czech, Dakota, English, French, German, Gikuyu, Italian, Japanese, Latin, Polish, Portuguese, Russian and Spanish, and smaller indicative fragments for Baoule, Dan, Dutch, Hua, Nyanja, Sanskrit, Serbo-Croat, Swahili, Tem and Welsh Romany. Anonymous FTP to ftp.cogs.sussex.ac.uk and directory /pub/nlp/DATR provides access to various DATR implementations, the example archive, and some relevant papers and documentation.

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

...coda

It is possible to further divide the onset, peak and coda to encode the distinct segments within them, but it is not necessary for our present purposes.

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

...follows

It is important to note that the definitions here are only default definitions which can be overridden for subclasses of words or individual words. Thus we make the assumption that the default structure of a root is a single syllable, but this is just a default assumption which will be overridden for any non-monosyllabic roots.

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

...n

Our German phonological segment inventory is taken from CELEX (Baayen et al., 1993) and uses the SAM-PA machine-readable phonetic alphabet (Wells, 1987).

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

...fragment

Verbal inflection in German does require reference to primary lexical stress and there are interesting stress shift effects in nominal inflection (see Bleiching 1992, 1994) but as it is not required for the inflection described here we have omitted a treatment of stress.

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

...output

It is a simple task to modify the rules we give so as to make the tree structure explicit in the way inflected forms are encoded.

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

...structure

Notwithstanding this expository simplification, we assume the (conventional) view that all syllables in polysyllabic roots have the same basic structure: we take a root such as Freude (/f r O y d @/) to consist of two syllables, with the second syllable having the onset /d/ and the peak /@/. Our view of polysyllabic roots is thus distinct from that of Bleiching (1992, 1994): she would represent the /d/ of Freude as the coda and the /@/ as the ``coda extension'' of the initial syllable. Our notion of focussed syllable thus equates with the only (conventionally structured) syllable in the disyllabic roots presented in Bleiching (1992, 1994). While the notion of focussed syllable was not discussed explicitly in either Cahill (1993) or in Bleiching (1992, 1994), it is implicit in this work.

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

...E

The details and consequences of this treatment of Der are spelt out in much greater detail in Section 8, below.

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

...morphology

Although we use different technologies, our general view on the proper relations between morphosyntax, morphotactics, and morphophonology is same as that outlined by Krieger et al. (1993, 146). The similarities and differences between the two approaches are discussed in detail in Evans and Gazdar (1996, 208-12)

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

...rules

Compare Zwicky's rules 3, 21#1, 7 and 11#1, respectively (1985, 381).

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

...adjectives

Bierwisch (1967, 256) cites lila (`lilac') and rosa (`pink') as instances of this class. All adjectives are uninflected when they appear as predicative adjectives.

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

...forms

The choice is apparently not determined by morphosyntactic features, declensional class or indeed any linguistic factors, but is, according to Hammer (1977,48), ``largely determined by ear''!

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

...forms

Viel and wenig also occur as members of syntactic categories other than determiner but they do not inflect at all in these occurrences. We are only concerned with their determiner instances here.

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

...follows

Zwicky's determiner classes as well as sharing inflection patterns also share syntactic information about the inflectional patterns of the adjectives with which they combine. We do not include this information in our fragment, restricting ourselves to inflectional information, but our classes map onto Zwicky's in the following way: Det_0 2#2 Det_I, Det_1A 2#2 Det_II, Det_2 2#2 Det_II, Det_3 2#2 Det_III. Our Det_1B class cannot be categorised in this way since the words in this class only appear in inflected forms when no adjective is present.

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

...instances

Jeder (`each/every'); solcher (`such'); jener (`that'); dieser (`this'); welcher (`which'); ihr (`your/her'); euer (`your'); unser (`our').

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

...coda

The node names here are the citation forms, not the roots, which are mostly monosyllabic.

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

...family

Dein (`your'); kein (`no'); mein (`my'); sein (`his/her/its').

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

...follows

The observant reader will spot that our morphology provides plural forms for ein. We make no apology for this. Although it is straightforward to augment the definitions so as to eliminate the plural forms, that would be a bad analysis. Ein is not morphologically exceptional in only having singular forms - it is entirely regular morphologically. However, it is inherently singular as a matter of syntax (and semantics) and thus has no use for its (regular) plural forms.

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

...form

This discussion is simply intended to show how compound determiners might be handled in the context of our theory of German inflection. It is not intended to constitute a general theory of compounding in German.

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

...German

Jenig is listed in Grimm's Deutsche Wörterbuch, Vol 4.2, 1877, and described as being derived from jener in the same way as einig is derived from ein.

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

...syntax

And we are grateful to one of our referees for drawing it to our attention.

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

...pronouns

We do not include a discussion of the first and second person personal pronouns in this paper. The patterns of variation are simply not interesting enough to warrant inclusion. And, unlike the third person pronouns, they are not related to the other inflectional patterns described here.

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

...forms

We implement this by pointing the relevant paths at the node UNDEFINED. This node is not given any definition in the fragment and so these paths also end up with no values defined. This is a standard DATR technique for characterising partial functions.

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

...forms

We have elected to leave all the genitive forms undefined. To the extent that putative genitive forms of German personal pronouns do occur, they are identical to the citation forms of the corresponding possessive determiners. But they do not occur freely and we incline to the view that such occurrences are really possessive determiner tokens, not genitive pronoun tokens.

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

...adjectives

Compare the account of essentially the same set of facts provided by Bierwisch (1967). Our account shares his concern for comprehensive coverage and analytical precision although the theoretical underpinnings of the two approaches are very different.

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