nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) (12/16/88)
NL-KR Digest (12/15/88 19:38:47) Volume 5 Number 38 Today's Topics: Re: shallow parsing (was Crass prejudice). Re: Verb-preposition idioms Submissions: NL-KR@CS.ROCHESTER.EDU Requests, policy: NL-KR-REQUEST@CS.ROCHESTER.EDU ---------------------------------------------------------------------- Date: Wed, 30 Nov 88 14:24 EST From: Greg Lee <lee@uhccux.uhcc.hawaii.edu> Subject: Re: shallow parsing (was Crass prejudice). From article <2052@garth.UUCP>, by smryan@garth.UUCP (Steven Ryan): > ... > Meanwhile, I continue my own research in parsing shallow with wide rules > instead of deep rules. In case the rather nasty tone of your recent postings is due in part to your feeling that the shallow parsing proposal did not receive the discussion it deserved, perhaps it would help if I say that I agree. For my own part, I just didn't think I had anything constructive to say. But in the interest of renewing that discussion, I'll make a few comments, anyhow. >[from earlier posting] ... >The suggestion was that humans memorise faster than generate and are better >at handling large chunks in parallel than small chunks nested. I think it's a plausible idea, and also it's implication that linguists' methodological reliance on economy of language description is mistaken. >What I take that to mean, we memorise words in each inflected form, even >regular forms, ... But an objection to this is that there is so much difference among languages in regard to what is expressed by inflections versus what is expressed using independent words -- periphrastically, that is to say -- without, however, any evidence for a corresponding difference in learning time or cognitive burden. > and possibly even groups words possibly up to words. .... Even for the case of regularly formed phrases? Not everything can be memorized. >Than language generation consists of inserting these chunks into a verb >frame chunk, with each sentence form being a different chunk. > >I think I'll call this suggestion shallow parsing: the parse tree will >only go down two or three levels before running into unanalysed (ig est >memorised) chunks. In terms of a productions, this would mean having >thousands of similar yet distinct productions instead factoring the >similarities to reduce the number of productions. If we disabbreviate the rules of a GPSG description as a set of context-free rules, it would look like this I think, at least in regard to the treatment of inflected forms. Greg, lee@uhccux.uhcc.hawaii.edu ------------------------------ Date: Wed, 30 Nov 88 17:55 EST From: Rod McGuire <mcguire@antares.aero.org> Subject: Re: Verb-preposition idioms The preceding discussion on "what particles really are" has not been all that illuminating, so I thought I would trot forth my opinions, namely: 1) Particles and words sometimes called locational/directional adverbs (e.g. "here", "home", "away", and "upwards") are really intransitive prepositions. 2) VPs containing transitive prepositions have at least 2 constituent analyses rather than a unique one of the form [v [p n]]. ARGUMENTS FOR ANALYSIS AS PREPOSITIONS 1) The set of words that can serve as transitive and intransitive prepositions form a closed class - one cannot make up new members. (while somewhat circular this is still an argument for them being assigned the same category). Some of these words can only be intransitive while others can also be transitive. Possibly, some can be only transitive. The best candidates are those words that are mainly structural and carry little meaning, e.g. "to", "for" and "of" - though proving this will be difficult. 2) Many verbs lexically subcategorize for prepositions to fill an argument in the propositional representation of the verb. And the verbs do not care whether this argument if filled by an intransitive preposition or by a transitive preposition and object. John went [home] | [to the store] | [out]. John put the dishes [on the table] | [away]. Categorically we can notate intransitive prepositions as P, transitive ones as P/N (i.e. a P missing a following noun or noun phrase), and verbs that need a preposition as V/P. Thus such a verb can either be satisfied by a P or the sequence [P/N N] which collapses to a P. 3) For people that don't think that "intransitive prepositions" are prepositions, there is the question of what the term "prepositional phrase" is supposed to mean. A few years ago, I was wondering (and still am) why linguists think the category VP exists (as distinct from S and V). I found Ruland Wells' 1949(?) article "Immediate Constituent Analysis" which was influential in defining the terms "noun phrase" and "verb phrase". There he defines X-phrases as phrases that can be substituted for a single word of category X in a "good" sentence to yield an equally good sentence. For example, starting with the sentence "men sleep" (= N V), the phrase "the old goats" can be replace "men" and thus it is a noun-phrase, and likewise "love Mary" can replace "sleep" and thus it is a verb-phrase. (BTW he punts on why VP should be distinct from S). I don't remember if he mentions prepositions, but by his criteria phrasal prepositions should be defined in terms of intransitive prepositions. ARGUMENTS AGAINST [V/P [P/N N]] AS A UNIQUE CONSTITUENT ANALYSIS Above I presented a categorical notation that allows the reduction: V/P P/N N | \ / V/P P \ / V However this notation also allows a constituent analysis of: V/P P/N N \ / | V/N N \ / V Or in other words, a transitive preposition can turn a verb that requires a preposition into a transitive verb. And there is a fair amount of evidence that this constituent structure is equally valid. 1) Incorporative verbs: There are a few verbs in English that appear to include prepositions internally, for example "entered" in: John [entered] | [went into] the room. Since "went into" is equivalent to a single word, it too must be a transitive verb. 2) Conjunction: [V/P P/N] phrases can be conjoined with transitive verbs: John [stepped on] and [broke] his glasses. 3) Passive: since [V/P P/N] phrases are transitive, passive forms can exist: John's bed has been [slept in]. 4) Parsing: In left-to-right parsing of text, one can consistently group [verb preposition] phrases as a constituent, without any look-ahead, regardless of whether the preposition turns out to be transitive or not. This allows idiomatic particles to be parsed in the same manner as "locational adverbs" and transitive prepositions. In fact, a parser may have to construct a [verb preposition] constituent just to look up whether or not the phrase is idiomatic and the preposition is intransitive. CONCLUSION I've discussed verb+prepositions only in terms of what constituent structures they can have. And I think that this is all that there is to syntax. In addition to simple sentences, there are various other phrasal forms (aka "transformations") in which particular verb+prepositions can or cannot appear. Judging the "acceptability" of these forms however is outside the realm of syntax. References that mention problem cases are: A.S. Kroch's review of Bruce Fraser's "The verb-particle combination in English" that appeared in Language V55 N1 (1979), and some paper I can't find but I think is by the Clarks or one of their students at Stanford that appeared in Language around 1980 on when [V/P P/N] phrases are fully transitive. Roderick McGuire - mcguire@aerospace.aero.org ------------------------------ Date: Thu, 1 Dec 88 20:35 EST From: mcguire@aerospace.aero.org Subject: Re: Verb-preposition idioms -------- [ I sent the following out over sci.lang, however our uucp mail connections seem rather disrupted now, so I'm sending this direct, on the off chance that I can still reach you ] From: mcguire@antares.aero.org (Rod McGuire) Newsgroups: sci.lang Subject: Re: Verb-preposition idioms Date: 30 Nov 88 22:55:39 GMT Reply-To: mcguire@aerospace.aero.org (Roderick McGuire) Organization: The Aerospace Corporation, El Segundo, CA Lines: 116 Summary: intransitive prepositions The preceding discussion on "what particles really are" has not been all that illuminating, so I thought I would trot forth my opinions, namely: 1) Particles and words sometimes called locational/directional adverbs (e.g. "here", "home", "away", and "upwards") are really intransitive prepositions. 2) VPs containing transitive prepositions have at least 2 constituent analyses rather than a unique one of the form [v [p n]]. ARGUMENTS FOR ANALYSIS AS PREPOSITIONS 1) The set of words that can serve as transitive and intransitive prepositions form a closed class - one cannot make up new members. (while somewhat circular this is still an argument for them being assigned the same category). Some of these words can only be intransitive while others can also be transitive. Possibly, some can be only transitive. The best candidates are those words that are mainly structural and carry little meaning, e.g. "to", "for" and "of" - though proving this will be difficult. 2) Many verbs lexically subcategorize for prepositions to fill an argument in the propositional representation of the verb. And the verbs do not care whether this argument if filled by an intransitive preposition or by a transitive preposition and object. John went [home] | [to the store] | [out]. John put the dishes [on the table] | [away]. Categorically we can notate intransitive prepositions as P, transitive ones as P/N (i.e. a P missing a following noun or noun phrase), and verbs that need a preposition as V/P. Thus such a verb can either be satisfied by a P or the sequence [P/N N] which collapses to a P. 3) For people that don't think that "intransitive prepositions" are prepositions, there is the question of what the term "prepositional phrase" is supposed to mean. A few years ago, I was wondering (and still am) why linguists think the category VP exists (as distinct from S and V). I found Ruland Wells' 1949(?) article "Immediate Constituent Analysis" which was influential in defining the terms "noun phrase" and "verb phrase". There he defines X-phrases as phrases that can be substituted for a single word of category X in a "good" sentence to yield an equally good sentence. For example, starting with the sentence "men sleep" (= N V), the phrase "the old goats" can be replace "men" and thus it is a noun-phrase, and likewise "love Mary" can replace "sleep" and thus it is a verb-phrase. (BTW he punts on why VP should be distinct from S). I don't remember if he mentions prepositions, but by his criteria phrasal prepositions should be defined in terms of intransitive prepositions. ARGUMENTS AGAINST [V/P [P/N N]] AS A UNIQUE CONSTITUENT ANALYSIS Above I presented a categorical notation that allows the reduction: V/P P/N N | \ / V/P P \ / V However this notation also allows a constituent analysis of: V/P P/N N \ / | V/N N \ / V Or in other words, a transitive preposition can turn a verb that requires a preposition into a transitive verb. And there is a fair amount of evidence that this constituent structure is equally valid. 1) Incorporative verbs: There are a few verbs in English that appear to include prepositions internally, for example "entered" in: John [entered] | [went into] the room. Since "went into" is equivalent to a single word, it too must be a transitive verb. 2) Conjunction: [V/P P/N] phrases can be conjoined with transitive verbs: John [stepped on] and [broke] his glasses. 3) Passive: since [V/P P/N] phrases are transitive, passive forms can exist: John's bed has been [slept in]. 4) Parsing: In left-to-right parsing of text, one can consistently group [verb preposition] phrases as a constituent, without any look-ahead, regardless of whether the preposition turns out to be transitive or not. This allows idiomatic particles to be parsed in the same manner as "locational adverbs" and transitive prepositions. In fact, a parser may have to construct a [verb preposition] constituent just to look up whether or not the phrase is idiomatic and the preposition is intransitive. CONCLUSION I've discussed verb+prepositions only in terms of what constituent structures they can have. And I think that this is all that there is to syntax. In addition to simple sentences, there are various other phrasal forms (aka "transformations") in which particular verb+prepositions can or cannot appear. Judging the "acceptability" of these forms however is outside the realm of syntax. References that mention problem cases are: A.S. Kroch's review of Bruce Fraser's "The verb-particle combination in English" that appeared in Language V55 N1 (1979), and some paper I can't find but I think is by the Clarks or one of their students at Stanford that appeared in Language around 1980 on when [V/P P/N] phrases are fully transitive. Roderick McGuire - mcguire@aerospace.aero.org ------------------------------ Date: Fri, 2 Dec 88 01:51 EST From: Mark Alexandre <alexandr@surya.cad.mcc.com> Subject: Re: Verb-preposition idioms In article <42261@aero.ARPA> mcguire@aerospace.aero.org (Roderick McGuire) writes: > > 2) Many verbs lexically subcategorize for prepositions to fill an > argument in the propositional representation of the verb. And the > verbs do not care whether this argument if filled by an intransitive > preposition or by a transitive preposition and object. > > John went [home] | [to the store] | [out]. > John put the dishes [on the table] | [away]. > Upon reading Mr. McGuire's well written analysis, it occurred to me that there is another form of conjuction that would seem to lend further credence to the analysis. Consider the sentence (with presumed partial bracketing): John went [home], [to the store], [[back] [home]], and [[out] for the evening]. Would Mr. McGuire or anyone care to comment on this? Furthermore, I am also curious about the anaylsis of the following sentences: Go on up ahead there. Stop on over by on your way home. Finally, during the discussion about the Germanic origins and parallels of this verb+preposition construct, I have developed a suspicion that the undisputed historical origin of the phenomenon in Germanic may be of very limited value in the synchronic syntax of Modern English. That is, I have an intuitive feeling that the structures may have undergone a reanalysis at some point in English and that only a remnant of the original Germanic form survives in Modern English alongside some other structure such as the one presented by Mr. McGuire. I have been looking for evidence and have come up the following scattered observations on the lexicon. 1) The number and the distribution of prefixed particles in English (along the mode of German) appears to be in decline. For example, I found in a pass over the Webster's Collegiate Dictionary: - no verbs beginning with the prefixed on- - only 2 verbs with prefixed by-: bypass and byline, both of which appeared within the last century and appear to be backformations derived from corresponding nouns - only 2 verbs with prefixed off-: offset and the hyphenated off-load, the latter attested only back to 1850 - only a few verbs with prefixed in-: -- inset, inweave, and inlay, all pertaining primarily to handicraft and handiwork; -- infold, which I have never heard colloquially; -- inbreed, which like outbreed, arose around the turn of the century, probably in the argot of horsebreeding; -- input, which like output, has only recently taken on a usage as a verb in the field of computers. The number of German verbs beginning with an-, bei-, aus-, ein- appear far more numerous. 2) The meaning of English verb+preposition is often not the same as that of a corresponding verb with the prepositional prefix. bypass vs. pass by offset vs. set off outwork vs. work out outkick vs. kick out overcome vs. come over overrun vs. run over overdo vs. do over undergo vs. go under uphold vs. hold up The English pairs are sometimes completely different in meaning. Sometimes the meaning of the prefixed form is included among several meanings of the verb+preposition form, suggesting that the former has stagnated while the latter continues to acquire new meanings as the language evolves. There is also the strong impression that the verb+preposition forms include far more colloquial and idiomatic meanings than do the prefixed forms. In German, the prepositional prefix, if detachable, is obligatorily detached in the tensed forms of the verb, but without any change in meaning. 3) Some prepositional prefixes are completely productive morphologically i.e. able to form new verbs with a predictable semantics. outdo, outperform, outlast, outshine, out-rehearse, outproduce, out-conceptualize, outkick, etc. overdo, overwork, overanalyze, overexagggerate, overproduce, overconcentrate, over-rehearse, etc. Notice that for these forms, there are never equivalent verb+preposition forms. Also notice that even non-Germanic root verbs may be so prefixed. These productive processes account for a great many of the English verbs of this form, again suggesting that those based on the original Germanic verbs are in decline. 4) Perhaps more importantly, some free prepositions are completely "productive", though it is certainly unclear whether this is a morpho- logical process or simply syntactic, as Mr. McGuire's analysis would suggest. Notably "by" and "over" seem to be able to complement any verb involving a resultant change of location: pass by, march by, swing by, pop by, stroll by, swim by, hop by, slither by, scoot by, jog by pass over, march over, swing over, pop over, stroll over, swim over, hop over, slither over, scoot over, jog over. The word "on" can be used to express the continuance of an action: go on, talk on, burn on, sail on, drag on, continue on, dream on, equivocate on, babble on. Are German bei- and ueber- similarly productive? What about equivalents in other Germanic languages? A lack of comparable productivity in the Germanic morphological process would certainly weigh heavily in favor of a syntactic analysis for Modern English verb+prepositions. This also ties in with the example sentence given earlier, in that it is just these usages that seem so readily combinable in series: pass on by, march on over, drop over by, babble on and on For all of these reasons, it just doesn't look to me like the "Germanic analysis" holds water when carried over to Modern English. The intransi- tive preposition analysis laid out by Mr. McGuire, however, appears much more promising. Comments and corrections are welcome, Mark Alexandre ------------------------------ Date: Sat, 3 Dec 88 08:44 EST From: Greg Lee <lee@uhccux.uhcc.hawaii.edu> Subject: Re: Verb-preposition idioms From article <42261@aero.ARPA>, by mcguire@antares.aero.org (Rod McGuire): " The preceding discussion on "what particles really are" has not been " all that illuminating, so I thought I would trot forth my opinions, "... It's an intriguing analysis. I'm very fond of combination by functional composition (V/P P/N = V/P) as opposed to functional application (V/P P = V). The dual analysis does give a neat way of describing passivizable objects of prepositions. (But what about the non-passivizable ones?) The earlier discussion centered around the extraposable verb "particles", which sometimes seem preposition-like ('he put the cat out') and sometimes not ('he looked the number up'). Does this categorical analysis offer any advantage in this regard? I don't see that it does. " A few years ago, I was wondering (and " still am) why linguists think the category VP exists (as distinct from " S and V). Beats me. *I* don't think VP exists. There was some discussion of the matter a while back, based on putative transformational properties of VP (do-so replacement, VP-preposing), without any clear evidence emerging, that I noticed. The Rulon Wells criterion you mention has the obvious drawback that it forces new phrases to be assumed wherever optional constituents are found, without any discernable rationale. Greg, lee@uhccux.uhcc.hawaii.edu ------------------------------ End of NL-KR Digest *******************