Auxiliaries: Difference between revisions

From Parallel Grammar Wiki
Jump to navigation Jump to search
(Created page with "In classic LFG (Bresnan 1982), auxiliaries were treated as subcategorizing for an open complement (VCOMP in those days, XCOMP in today's terms). See Kaplan and Bresnan (1982:2...")
 
(Blanked the page)
Tags: Blanking Visual edit
Line 1: Line 1:
In classic LFG (Bresnan 1982), auxiliaries were treated as subcategorizing for an open complement (VCOMP in those days, XCOMP in today's terms). See Kaplan and Bresnan (1982:205-206, 228) for concrete examples. This basic approach is laid out in some detail in Falk (1984).


In some of the earliest discussions within ParGram, it was realized that while the XCOMP analysis for auxiliaries has some useful consequences in English such as being able to deal with the affix dependencies among auxiliary stacks in English quite easily (cf. Chomsky's notion of Affix Hopping) and has the advantage that coordinations like "John will and can disappear." are no problem, there are also major disadvantages from a crosslingusitic perspective. In languages like German or French (English, French and German were the original ParGram languages), some tense information is encoded via auxiliaries whereas some other information is encoded via inflections on the verb. Butt, Niño and Segond (1996/2004) use the following example sentences in the future perfect to demonstrate the problem:
The driver will have turned the lever. Der Fahrer wird den Hebel gedreht haben. (German) Le conducteur aura tourné le levier. (French)
In the English version, the future part is expressed by "will", the perfect part by "have" in combination with the past participle of "turned". In the German version, the picture is similar, except that there is a different word order. In French, however, the future and the perfect are combined into one auxiliary: "aura".
Under an XCOMP analysis, this means that the English and the German will have two layers of XCOMP embeddings, while the French will have only one. In terms of parallelism across languages (and back then we had machine translation as our main application goal in mind), this is not an acceptable way of proceeding. Butt et al. instead proposed a "flat" structure in which TNS/ASP information across languages is registered within a TNS/ASP feature, but in which auxiliaries (and other functional elements) are not analyzed as embedding event arguments (which is what XCOMPs are).
The particular way this was implemented in 1996 was via m(orphological)-structure (see [[m-structure]]). The problem was what to do about morphological wellformedness information like the affix dependencies found in English. These clearly need to be taken care of, however, they are not interesting for any further analysis as they tend to be quite language specific. That is, f-structure is not an appropriate level to represent language-particular morphological information that is only used to ensure morphosyntactic wellformedness. The ParGram grammars experimented with m-structure for a short while (see pages 60-67 in the Grammar Writer's Cookbook), but then abandoned it as too unwieldy (debugging becomes very difficult whenever several different projections from c-structure must be kept track of). Instead, the ParGram grammars now store information needed for language particular morphosyntactic wellformedness checking in a [[CHECK feature]] at f-strucutre. All the information bundled there can be removed/ignored for further analysis steps such as semantic analysis, machine translation, etc.
However, within theoretical LFG, m-structure took on a life of its own and current work continues to explore uses or ramifications of adding in this further projection. See the [[m-structure]] page.

Revision as of 10:27, 12 January 2023

Retrieved from "https://wiki.uni-konstanz.de/pargram/index.php?title=Auxiliaries&oldid=10"