1 / 24

Faithfulness, Registers and Lexical Representation

Faithfulness, Registers and Lexical Representation. Marta Castella°* - Marko Simonovi ć * (° Università di Verona, *Utrecht University – UiL-OTS). Verona IGG38 – February 24 th. Structure of this talk. Starting point : Van Oostendorp (1998 )’s idea Data and Analyses

hashim
Download Presentation

Faithfulness, Registers and Lexical Representation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Faithfulness, Registers and Lexical Representation Marta Castella°* - Marko Simonović* (°Universitàdi Verona, *Utrecht University – UiL-OTS) Verona IGG38 – February 24th

  2. Structure of this talk • Startingpoint: Van Oostendorp (1998)’s idea • Data and Analyses • Roman dialect • Dutch • Serbo-Croatian • Consequences • Our contribution to the model • Residual Questions

  3. Starting Point Van Oostendorp (1998) works out the idea that styles and registers of one language (informal, formal, etc. – and we include speed as well) represent related systems and not self-contained grammars. This is to simplify the set of assumptions that we need on the acquisition side (avoid an “empowered” logical problem of language acquisition). make inconsistent that two registers would differ like two random grammars do (on the empirical/typological side). We produce extra evidence in support of this claim, using data of slow, medium and fast speech from the dialect of Rome and from Serbo-Croatian.

  4. Theoretical framework OT Prince and Smolensky (1993/2004): • Surface forms are the result of the evaluation by conflicting constraints that interact with one another, dependently on their ranking (a language-specific hierarchy); • a grammar consists of the ordering of this universal set of constraints, and the way of accounting for any variation is the manipulation of their ranking. Van Oostendorp (1998) suggests a principle to account for cases of register variation: • The more formal the register, the higher ranked the faithfulness constraints

  5. Example 1 Allegroform: speed and the dialect of Rome In the dialect of Rome, speed has a clear impact on phonology. From slow to fast speech many phenomena arise. One is, for instance, the deletion/lenition of [l] followed by vowel assimilation. We suggest that: • formal levels behave symmetrically to slow speech, • semi-formal levels behave symmetrically tomedium speech • informal levels behave symmetrically to fast speech There is a continuum between perception-oriented speech (i.e. listener-friendly) and production-oriented speech (i.e. speaker-friendly).

  6. Deletion of [l] followed by vowel assimilation Prepositions: /de/ “of, from” /da/ “from, to” /a/ “to” Articles:/lo/* “themasc.sing”, /la/ “thefemm.sing”, /i/“themasc.plur”, /e/“thefemm.plur” Data Sample [slow] [de lo] [de la] [a la] [a lo] [medium] [de.o] [de.a][a.a][a.o] [fast] [doo] [daa][aa][ao] [slow] [da lo] [da la] [da le][da i] [medium] [da.o] [da.a] [da.e] [da.i] [fast] [doo] [daa] [dee] [dii] */lo/ is the article that precedes words starting with /z/, /pn/, /ɲ/, /ps/, /ʃ/, /ks/, /s/+consonant(and vowels). Elsewhere /er/.

  7. Constraints and Rankings Constraint set Markedness • *(weak)struc: weakconsonants are notpresent in the output • Agree: adjacentvowelshave the samefeatures Faithfulness • IDENT-V- IO: Itisprohibited to change the featurevalues in vowels • MAX-C-IO: No deletion in consonants Ranking: [Slow] IDENT-V-IO, MAX-C-IO>>*(weak)struc, Agree

  8. Outcomes

  9. Constraints and Rankings Constraint set • *(weak)struc: weakconsonants are notpresent in the output • Agree: adjacentvowelshave the samefeatures • IDENT-V- IO: Itisprohibited to change the featurevalues in vowels • MAX-C-IO: No deletion in consonants Ranking: [Med.] *(weak)struc, IDENT-V-IO>>MAX-C-IO, Agree

  10. Outcomes

  11. Constraints and Rankings Constraint set • *(weak)struc: weakconsonants are notpresent in the output • Agree: adjacentvowelshave the samefeatures • IDENT-V- IO: Itisprohibited to change the featurevalues in vowels • MAX-C-IO: No deletion in consonants Ranking: [Fast] *(weak)struc, Agree>>IDENT-V-IO, MAX-C-IO

  12. Outcomes

  13. Reranking [Slow] IDENT-V-IO, MAX-C-IO>>*(weak)struc, Agree [Med.] *(weak)struc, IDENT-V-IO>>MAX-C-IO, Agree [Fast] *(weak)struc, Agree>>IDENT-V-IO, MAX-C-IO

  14. Reranking [Slow] IDENT-V-IO, MAX-C-IO>>*(weak)struc, Agree [Med.] *(weak)struc, IDENT-V-IO >> MAX-C-IO, Agree [Fast] *(weak)struc, Agree>>IDENT-V-IO, MAX-C-IO

  15. Example 2 (from van Oostendorp 1998) Vowel reduction in three registers of Dutch fonologie (phonology) is pronounced: • [ˌfo.no.lo.ˈxi] formally • [ˌfo.no.lə.ˈxi] semi-formally • [ˌfo.nə.lə.ˈxi] informally Constraint set • Parse (F) • Reduce-1:Weak and semi-weak positions should be schwa. • Reduce-2: Weak positions should be schwa. Immediatelyfollowing an unstressed position Allother unstressed postions

  16. Example 2:Rankings and outcomes • formal. [fonoloˈxi] yields [fonoloˈxi] Parse-[+high] >> Parse-[+round] >> Parse-[+low] >> Reduce-2 >> Reduce-1 >> Parse-[+front] • semi-formal. [fonoloˈxi] yields[fonoləˈxi] Parse-[+high] >> Reduce-2 >> Parse-[+round] >> Reduce-1 >> Parse-[+low] >> Parse-[+front] • informal. [fonoloˈxi] ] yields [fonələˈxi] Parse-[+high] >> Reduce-2 >> Reduce-1 >> Parse-[+round] >> Parse-[+low] >> Parse-[+front]

  17. New data – Example 3Opacity, Coda-l vocalisation and a-Epenthesis in Serbo-Croatian In Serbo-Croatian, there is a historicized process which turns coda-l into [o]. Coda-l vocalisation /del/ → [de.o] “partnominative” /del + a/ → [de.la] “partgenitive” (It does not apply to any non-native items: kanal, rival, interval.) There is another process which disrupts all coda clusters by inserting an epenthetic [a]. a-Epenthesis /visk/ → [visak] “pendulumnominative” /visk + a/ → [viska] “pendulumgenitive”

  18. Example3 - Opacity and Registers These two processes interact opaquely: Coda-l vocalisation bleeds the environment of a-Epenthesis. /petl/ → /petal/ → [petao] “cocknominative” (a-epenthesis) (l-vocalisation) /petl+ a/ → [petla] “cockgenitive” Important! Fast/very informal has the transparent mapping /petl/ → [peto].

  19. Ourcontribution Potential problem: Registers are rankings of constraints i.e. OT grammars. However, informal/faster registers could, in principle, be unlearnable by themselves because, as we have seen, they rely on representations which are made available by formal/slower registers. Solution: Other representations another language. So registers are full rankings, butnot full, self-sufficientlanguages. Generalization: registers are part of a broaderphenomenon of relatedforms (e.g. speed, dialects, cognates, interspeakervariatios, etc.) AcquisitionAccomodationPrinciple: when speakers are building theirlexicon/grammar, theyrely on all the availableforms. Bottomline: representations matter (not contraRotB, rather proRotS)

  20. In other words, a direct consequence of the data: what if a register is missing? • What is the representation in a Dutch speaker who has only ever been exposed to [ˌfonələˈxi]? But then, this speaker would have to be exposed to [ˌfonəˈlox] “phonologistinformal”? • Can there be a Roman fast-register-only speaker? Would this speaker lack any representation of [de la]? • How does the Serbo-Croatian “informal speaker” compute the opaque forms of the type [petao]? Bottomline: Representations depend on the available forms.

  21. Beyond the primacyof the slow/formal? • Whatwehaveseenisreally part of the Dutch/Italian/Serbo-Croatian speakers’ competences. • So itisworthwhile to movebeyond the “anthropological” bias(encoded in van Oostendorp’sprinciple): Slow/formalgetsit right by definition: the lexiconrelies on slow/formalspeech. So, itisnotsurprisingthat FAITH is high in slow/formalranking. • But: do weneed to encodethatmuch?

  22. Our alternative or On how van Oostendoropis (more) right (than he thinks) • Every surface form (produced or perceived) participates in the construction of the underlying representation. • This clearly plays a role in the comprehension of e.g. other dialects/sociolects (see Benders 2011). • Due to physical properties of slow/formal speech, such forms will contain more contrast/information than other forms; these will always be closer to underlying forms.

  23. Residualquestions As pointed out to us by Birgit Alber (voce), this could be a domain of Output-Output relations. So, do different style/register forms constitute paradigms of some sort? • Is there cross-style uniformity? • Is there blocking from above? • e.g. in lexicalizations, borrowings, etc.

  24. Thankyou! m.castella@uu.nl rkicma@gmail.com

More Related