Ryan Georgi Ling 567 Lab 6 Writeup -----========= Discourse Status: ==========--------- Pt. I: DEFINITES --------------- Getting at discourse status is difficult from what I've learned about Arabic so far, but I'll try to understand it by writing it down as I go. To begin with, something that seems to "anchor" what I've read so far seems to be that the Demonstratives, which logically should be active-or-more, when used adjectivally, must have the head NP grammatically definite (this means one of three things; either marked with the definite prefix /?al-/, marked with a pronomial clitic (suffix) used for possessives, or if the following noun is in the genitive (for "construct" phrases). This would seem to indicate to me that definiteness must be at least partially correlated to whatever status is aligned with demonstratives, which I will get to later, but seem to be activ-or-more. Some more info that helps place the definite marker -- As noted by the holes text, UniqueID cases are marked in this way. e.g. ?al-qamar (the moon), ?al-wila:ya:tu l-muttah.ida (the United States) (p. 196) There are other examples where the focus seems to be closer to active-or-more: ?al-yawm = Today (lit. "the day") ?as-sa:'a = at the time (lit. "the hour") (p. 196) Lastly, generic refernce (though we're calling it outside the scope of the class) seems to pick out the familiar status, with: qami:s. mina l-qut.n shirt of the-cotton 'A cotton shirt' Lastly, when used with any time of verbal nouns or nominal adjectives the familiar case seems to again be used: ?al-tadxi:n xat.ar 'ala: s.-s.ih.h.a the-smoking danger to the-health 'Smoking is dangerous to the health' (p. 198) Taking all of these cases into account, it would seem to me that the /?al-/ prefix denotes a uniq-or-more status for the nouns it marks -- a pretty broad scope, but one that accounts for all these various uses. Thankfully, since I had implemented the lexical rules for these markings in lab 5, for this lab I simply added the COG-ST constraints onto the existing rules. Pt. II: INDEFINITES ------------------------ From what little info I was able to surmise about definite marking, there is little or no information given on the final /-n/, which I have come to learn is often mistakenly thought of to be an indefinite marking. According to wikipedia (http://en.wikipedia.org/wiki/Nunation), Nunation is slightly different from an indefinite marker; it more precisely marks that the noun is NOT marked for definiteness. What seems to be confusing about this is that this final /-n/ is not usually written in the orthography except for the feminine accusative, and is only pronounced non-"pausally", which is a form of pronunciation apparantly mostly associated with formal reading aloud. Pausal rules require the nunation to be pronounced only prepausally, (before any sentence-ending punctuation). It appears that there are several exceptions, but in general, then the /-n/ final form is the "default" marking for a noun, and is removed when the definite is attached. Any time it is missing in the examples I've found is probably due to the above restrictions on pronounciation and lack of presence in the original orthography. As a result, ultimately, also due to the large scope shown by the ?al prefix, it seems that the -n ending should be constrained to the type-id case. I added this to my pre-existing lexical rule for indefinites. ------========== DEMONSTRATIVES ========-------------- Proximal Distal Sg. Masc. ha:Da: Da:(li)ka Fem. ha:Di(hi) tilka Pl. Both ha:?ula:?i ?u:la:?ika M. Nom. ha:Da:ni Da:nika Dual Acc/Gen ha:Dayni Daynika F. Nom. ha:ta:ni ta:nika Acc/Gen ha:tayni taynika ...these forms are marked for gender, number, and case, though the latter only shows up in the dual forms. ha:- is the Proximal marker, while -(li)ka marks distals. Since the forms are so close, I will be adding the "base" forms in, then adding whether they are proximal or distal with a lexical rule. (Making the singular feminine an exception.) It appears that the demonstrative pronouns in Arabic have three usages; as adjectives and presentatives (a form of emphasis, specifically for present objects), and as subjects of "verbless" copulas. This second case appears to be the only way they are ever used as a pure pronoun. bayt-i: ha:Da: ha:Da: bayt-i: house-my this this house-my 'This house of mine' 'this is my house' * As an interesting note, when the following noun has been made definite but Since I have not implemented the copula phrase yet, and since this is the minority case, I will choose to implement the adjectival version discussed below. In all the adjectival cases, demonstratives occur with grammatically definite NPs. In the case of the simple definite, they precede the noun, while in cases of either posessive pronoun affixation or the genitive "construct" phrases (see example below), they occur after. ha:Da: l-?akl ?akl-i: ha:Da: this the-food food-my this 'This food' 'This food of mine' I haven't yet gotten to this case, so for now I'm just dealing with the raw adjective by giving it its own new form, demon_adjective (scary!) and having it constrain the element on its COG-ST to be activ-or-more. ----------========== DITRANSITIVES ==========-------------- I had a terribly hard time trying to find a description of ditransitive verbs in Arabic. My grammar has no chapters describing it, and searching the web for "ditransitive" in arabic doesn't produce much. Furthermore, it appears from what I have gleaned that ditransitives have their second argument marked just by word order, and the term "indirect object" appears to not be helpful either, since for all other respects, this second argument appears just like the first object. The form appears to be: Where the subject may either be prefixed or suffixed, and the both objects may either be incorporated pronouns, or full NPs, with the "indirect" object being the first. ** My primary source for examples of this comes from my "Source C" examples; but as they use the past tense, I've commented them out in order to keep tense standard. -------============ ARGUMENT OPTIONALITY ======-------------- I searched for quite some time on this subject, and only encountered papers covering that which I already learned; that Arabic has rather systematic subject pro-drop and object incorporation as pronouns, but I found nothing concerning the possibility for the type of Discourse related object drop we're interested in. As far as I can tell, subject prodrop is allowed with definite instantiation, and object prodrop is not allowed; it must be made up with pronoun incorporation (where the COG-ST is made explicit by the pronoun). ARGUMENT PATTERNS: ------------------- I've found the following patterns, and tried to represent them all in my modified test suite. FOR ALL NOUNS: 1) No missing arguments, verb shows agreement with subject NP (case-marked) 2) Missing subject NP, verb shows agreement EXAMPLES: yatay rrajulu lwalada lkita:ba y-atay ?al-rajul-u ?al-walad-a ?al-kita:b-a 3MSG-give the-man-NOM the-boy-ACC the-book-ACC 'The man gives the boy the book' VERSUS: yatay lwalada lkita:ba y-atay ?al-walad-a ?al-kita:b-a 3MSG-give the-boy-ACC the-book-ACC 'He gives the boy the book' ...this latter case seems to be constrained to dropping the subject for activ-or-more. Furthermore, this seems to happen for all verbs, so I added OPT-CS activ-or-more to the verb-lex's SUBJ constraints. Since the verb ALWAYS shows agreement marking, the top two should be considered variants of all the following: -- Transitive: 3) Explicit object (generally only third person) 4) Incorporated objected pronoun EXAMPLE OF 4: y-adribu-hu ?al-rajul-u 3MSG-hit-him the-man-NOM ' The man hits him ' (would be 'he hits him' when missing explicit subj) -- Ditransitive: 5) Explicit (indirect) object, explicit (direct) object 6) incorporated (indirect) object, explicit (direct) object 7) incorporated (indirect) object, incorporated (direct) object ?? 8) Explicit (indiret) object, incorporated (direct) object Example of 5): yatay rrajulu lwalada lkita:ba y-atay ?al-rajul-u ?al-walad-a ?al-kita:b-a 3MSG-give the-man-NOM the-boy-ACC the-book-ACC Example of 6): yatayhu rrajulu lkita:ba y-atay-hu ?al-rajul-u ?al-kita:b-a 3MSG-give-him the-man-NOM the-book-ACC Example of 7): yatayhuhu rrajulu y-atay-hu-hu ?al-rajul-u 3MSG-give-him-it the-man-NOM ' The man gives him it' ... I haven't seen any examples about the grammaticality of an incorporated second argument (direct object) but an EXPLICIT first argument (ind. obj) , but I have found in snippets that suggest the dialects rely on word order for object understanding, so it seems there may not be anything I can do to block this directly, since it will look exactly like #6. UPDATE: I talked to Margalit about Hebrew, and she says in the case of double-object verbs in Hebrew, at least, there doesn't seem to be a clear distinction about how to do it one way or another, and it's largely dependent on semantics (i.e. giving a boy to a book?) (see http://www.ethnologue.com/show_language.asp?code=ary) TESTING: ---------------------- I added a bunch of instances of pronoun incorporation to my test suite, and I think I have most if not all cases covered. I tested: Broken: I haven't investigated thoroughly, but it seems when I try incorporating both the direct and indirect object ------------=========== MODIFICATION =============----------------- ADJECTIVES: ----------- As I didn't implement agreement in the last assignment, there was plenty of work for me to do here. First and foremost, was to get rid of the "FNT" head feature I'd introduced in an earlier lab, now that this distinction is subsumed by the COG-ST. I'd already implemented adjectives as posthead modifiers in lab 5, so the difficult part was simply getting agreement working. As I've mentioned, adjectives (with the exception of demonstratives) are post-head ( a fact that inspired me to create both posthead-adj and prehead-adj types for each of these to inherit from ) and agree with their nouns in case, number and definiteness. Since adjectives are inflected to match their nouns, I made lexical rules that added the appropriate affix for both nouns and adjectives; though for adjectives I needed separate rules so that I could add the constraints to the MOD feature of element the adjective selects for, since these features don't exist in an equivalent place on the adjective itself. TESTING: I no longer get parses when the case or definiteness of the adjectives do not match, and I believe most or all of the cases in which they do. ** Something I still haven't figured out, however, is when generating I get extra parses which seem to take definiteness and case as optional on adjectives, but not nouns. ADVERBS: ---------- Since both verbs and noun are often strongly related in Arabic through the consonantal clusters they all derive from, "verbal" nouns are quite common, and as such, from what I've determined, most productive adverb classes are actually adjectival in nature. Another interesting example is where a verbal noun is taken as an object to a verb, and then modified adjectivally. p. 257 in the Clive Holes book: darabahu darban s^adi:dan darab-a-hu darb-a-n s^adi:d-a-n hit-3msg-him hitting-ACC-INDEF strong-ACC-INDEF 'he hit him hard' This is the absolutive accusative construction, and rather uncommon. There is a relatively closed class of "pure" adverbs, however, that usually signal time and a number of different types of manner. No matter what time of adverb, the sources I've read seem to indicate that in the vast majority of cases the adverb simply attaches to the right (...in Latin orthography... that could get confusing!) of the entire VP, such as the example above, or: y-ins.arifu ?al-rajul-u kawam 3MSG-leave the-man-NOM quick(ly) 'the man leaves quickly' To make things simple until I can get into these verbal nouns more in-depth, I simply added this type "kawam" to my grammar and made it POSTHEAD +. With the rather simple sentences that I'm still parsing, this seem to be all that was needed to get kawam appearing in the right place. WRAP-UP: --------- I tried to fix a number of examples that were broken in my test suite. By introducing agreement as well as a few more lexical entries, I was able to seriously reduce overgeneration and increase coverage. On the examples I covered last time, I reduced the number of readings by half generally, and increased coverage (over the newly added material as well, admittedly) to now 26%.