150 likes | 152 Views
A paradigmatic account of lexical innovation: the case of Romance N+N compounding. Jan Radimský University of South Bohemia ( Czech R epublic ). Introduction. Aim : to e xamine and refine hypotheses accounting for the emergence of the subordinate N+N compounds (SUB- NNs ) in Romance
E N D
A paradigmatic account of lexical innovation: the case of Romance N+N compounding Jan Radimský University ofSouth Bohemia(Czech Republic)
Introduction • Aim: to examine and refinehypotheses accounting for the emergence of the subordinate N+N compounds (SUB-NNs) in Romance • French examples • exposition photos “photography exhibition” • abonnement internet “Internet subscription” • timbre-poste“postage stamp”, lit. “stamp-post” • sauce-tomate“tomato sauce” • Initialassumptions • The emergence of SUB-NN compounds in Frenchcan be accounted for in terms of Construction Grammar: a case ofconstructionchange(Closs-Traugott, Trousdale, 2013) • SUB-NNs are extensionsofAttributive NNs(such as bourgeoisgentilhomme– “bourgeois gentleman”) thathavealwaysbeenpresent in French(Radimský, 2019) • Bothpatternsgrow up exponentially during the 2nd half of the20thCentury • The SUB-NN pattern does not seem to have any restrictions nor preferences as for the semantic relationship between N1 and N2 (Arnaud2003:64) • Research question: • Does the family-size effect of repeated forms on either the N1 or N2 position have any relevance for the emergence or the conventionalization of the NN pattern? • Plan • A brief outline ofinitial assumptions • Data analysis: Frantext, FrWac • Type frequency spectra of ATTR and SUB NNs
Emergence of SUB-NNs (1) Rainer-Buridant (2015:1978)
Left-headed NN construction [NiNj]Nk↔ [Ni-headNj-non-head]k SUBNNs (timbre-poste) [NiNj]Nk ↔ [SEMiREL SEMj]k COORD NNs (boulangerie-pâtisserie) • [NiNj]Nk • ↔ • [SEMiis a SEMj]k • ↔ • [SEMjis a SEMi]k Emergence of SUB-NNs (2) ATTRNNs (bourgeois gentilhomme) • [NiNj]Nk • ↔ • [SEMiis a SEMj]k • The three NN constructions are instantiations of a more general left-headed NN construction (Radimský, 2019) • SUB NNS: “nonconventional extensions” of ATTR and COORD NNs • ATTR NNs: always available, no rival pattern • The attributive “is a” relation is extended to any semantic relation “REL” • ... [the] tolerance for nonconventionality is of great importance in change: partially sanctioned extensions of an existing conventionalized construction may over time become fully sanctioned instances of a more general, schematic construction, which has changed as a result of the speaker/hearer’s experience with language (Closs-Traugott, Trousdale, 2013:16) • Empirical support: theproductivityof the three subtypes are related (both token & type frequency curves in Frantext)
RelativetypefrequencyofNNs(Radimský, 2019) • Stable for over 100 years, an increase since the 1960s • Fq. of SUB-NNs follows the fq. of ATAP+COORD-NNs
Process of the construction change... • Role ofsemantics of the REL in theconstructionchange? • ATTR NNs = 1 relationship: [NiNj]Nk ↔ [SEMiis a SEMj]k • SUB NNs = ... 54 different semantic relations! (Arnaud, 2003:64) • ...no apparent regularity, any semantic relationship is possible • Role of repeated forms (N1, N2)? • ATTR NNs • Typical N2s with a high type frequency in Italian (Baroni, Guevara and Pirrelli, 2009; Grandi, 2009; Radimský, 2015), maybeeven“close set” of N2s (Grandi, 2009): chiave(“key”), base (“base”), modello(“model”), simbolo(“symbol”), vittima(“victim”) • [Nisimboloj]Nk ↔ [SEMiis aSEM(simbolo)j]kluogo simbolo ([place symbol], “symbolic place”) • Rationale: their specific meaning (expresses “quality”) triggers the attributive interpretation • SUB NNs • Surprisingly also some typical N2s with a high type frequency:sicurezza– “security”, stampa– “press”, lavoro– “work” (Baroni, Guevara and Pirrelli, 2009) • Koga (2018): N1s and N2s with a significant type frequency in French SUB NNs • No theoretical rationale? • Hypothesis: • If a noun features prominently in a NN pattern, it could more easily reappear in new NNs, irrespective of the semantic relationship between N1 and N2.
Data • Frantext Corpus (lemmatized section) • French fiction,127M positions, data from 1830 to 1999 • 33,800 binominals extracted (lemmatized types) • 1.631 NNs identified (353 SUB and 1,278 ATTR+COORD) • FrWaccorpus • French web corpus, 1600Mpositions • 430,000 binominals extracted (lemmatized types), token fq>3 • 3.350NNs identified (2,112 SUB and 1,238 ATTR) • Extremely low “signal-to-noise ratio” in both corpora • Manual identification of NNs is necessary
Type frequencies in Frantext(“old SUB NNs”) • Based on 152 “old” SUB NNs • Max. average year of appearance: 1959 • Type fq. spectrum of N1s • 85% of N1 s have type fq. 1 • Type fq. spectrum of N2s • 84% of N2s have type fq. 1 • Extremely few nouns with higher type frequency
Type frequencies in FrWac – ATTRBased on 1238 ATTR NNs • Type fq. spectrum of N1s • 67% of N1s have type fq. 1 • Type fq. spectrum of N2s • only 30% of N2s have type fq. 1 • one N2 (clé“key”) yields 18% of types (220 types) • Interpretation • Data confirm commonly shared assumptions about typical N2s with a high type frequency that “trigger” the ATTR interpretation • The opposite is observed for N1s: type frequencies per item are very low
Type frequencies in FrWac – SUBBased on 2112 SUB NNs • Type fq. spectrum of N2s • Type fq. spectrum of N1s • Interpretation • There are “typical” N1s and N2s with a high type frequency • Effect of repeated forms isbalanced, thoughslightly higher for N1s • Comparing the rate of Ns with type fq.=1 in different frequency spectra: • As the vocabulary of SUB NNs grows, there is more “regularity” based on repeated N1s or N2s • Is semantics involved in any way?
Stages in theconstructionchange (1) • SUB NNs as a non-conventionalextensionof ATTR NNs • General schematicpattern: [NiNj]Nk↔[Ni-head Nj-non-head]k • No further regularity, but new types seem to appear more frequently in specific contexts (business, marketing, management, IT...) • Creation of semi-schematic constructions based on a specified N1 or N2 such as papier (“paper”) or client (cf. Koga, 2018) • [[papier]iNj]Nk ↔ [SEMi REL SEMj]k • [Ni[client]j]Nk ↔ [SEMi REL SEMj]k • The relationship REL is underspecified, the same N is not necessarily associated with the same REL, cf.: • papier journal (newsprint = low-cost paper commonly used to print newspapers, lit. “newspaper paper”) • papier aluminium (aluminium foil, lit. “aluminium paper”) • papier toilette (toilet paper) • papier cadeau(wrapping paper, lit. “gift paper”) • BUT: some Ns (more frequently N1s) display typical RELs • e.g.: rayon+N2 (“department”) – refers to internal organization of a department store in sections • rayon boucherie(“meat counter”), rayon jouets (“toys department”), rayon enfant (“children’s section”), rayon [fruits et légumes] (“[fruits and vegetables] department”), rayon surgelés(“[frozen food] section”) • Notice, however,that the “sections” are referred to from different perspectives, so that N2 may denote goods(“toys”), targetcustomers (“children”) or names of respective stores (“butcher’s shop”), which would imply different RELs in a formal semantic analysis • RELisspecified, but ithas to be understood rather widely: • [[rayon]iNj]Nk ↔[sectionof a department storeidentified bySEMj]k • Such semi-schematic construction has an open-choice slot: [[rayon]iNj]Nk
Stages in theconstructionchange (2) • New schematic constructions emerge... • ... as abstractions over a set of semi-schematic constructions with similar REL: • rayon+N2 (“department”) – section of a department store (type fq. = 63) • pôle+N2 (“center”) – organization specialized in one activity (type fq. = 82) • atelier+N2 (“workshop”) – workshop specialized in one activity (type fq. = 42) • service+N2 – service specialized in one activity (type fq. = 40) • rubrique+N2 – (“column”) newspaper section (type fq. = 35) • filière+N2 – (“industry”) industry specialized in one activity (type fq. = 22) • coin+N2 – (“corner, area”) area of a house or store dedicated to an activity (type fq. = 25) • Consequence: the semantics (=REL) matters, the slot on N1 becomes open-choice • [NiNj]Nk ↔ [specific subdomain of SEMi related to SEMj]k • Other possible N1s: branche(“branch, field”), espace (“space, area”), point (“point”)... • What does “open-choice” mean here? • According to the most general construction, any left-headed NN structure is possible, but in many cases still non-conventional. • If the new NN type conforms to an intermediate semi-schematic (stage 2) or schematic (stage 3) construction, it is likely to be perceived as conventional.
Quantitativeaspects... • Semi-schematic constructions (stage “2”) • Hundreds of constructions based on lexically filled N1 or N2 • FrWac type fq.>10: 32 N1s and 40 N2s • Various degree of semantic regularity (REL) in each pattern • Schematic constructions (stage “3”) • Only very few cases for the time being • Semantic generalizations do not seem to be predictable • A set of semi-schematic constructions based on similar REL is necessary • Examples: • [NiNj]Nk ↔ [meal SEMi with taste/ingredient SEMj]k • NNs: sauce tomate(“tomato sauce”), arômevanille(“vanilla flavour”) • N1: sauce (22 types), arôme (10)... low type fq.: tarte, flan, fondant (= types of dessert), yaourt... • Common N2s for sweet meals help to fix the pattern • [NiNj]Nk ↔ [event SEMi dedicated to SEMj]k • NNs: soirée concert (“party+concert”), pause café (“coffee break”) • N1: soirée (“party”, 36 types), pause (“break”, 14) • N2s do not overlap – probably still not a vital pattern
Results & Discussion • Accounting for the emergence of French SUB NNs • Constructional change (= stage 1) • ATT NNs construction generalized to SUB NN[NiNj]Nk ↔ [SEMi-headis a SEMj-non-head]k [NiNj]Nk↔[Ni-headREL Nj-non-head]k • Any left-headed NN is possible, but most of them are non-conventional (“tolerance for nonconventionality”, Closs-Traugott, Trousdale, 2013:16) • Conventionalization of single substantive micro-constructions • E.g.: centre-ville (“town center”), appareil photo (“camera”), version papier (“print version”), timbre-poste (“stamp“) • Constructionalization (bottom-up process, = stages 2 and 3) • Repeated forms on either N1 or N2 position in micro-constructions lead to the creation of semi-schematic constructions based on a specified N1 or N2 • Type frequency spectra change: the rate of N1s and N2s with the lowest type fq. decreases • The semantics of the REL does not seem to beso important: REL maybeunderspecified • Abstraction of new schematic patterns • Generalization over sets of semi-schematic constructions with similar REL • French: the “sub-categorization” pattern • Italian: Verbal-nexus compounds (trasportopassegeri– “passenger transport”) • The process seem to be only very recent in Romance • The generalization may be based on any REL: it is not predictable • Specific constructions (2.1.+2.2.) may yield conventional NNs • Future work: How do new semi-schematic (2.1.) and schematic (2.2.) constructions interact with the N-PREP-N pattern?
References • Arnaud Pierre J. L. (2003), Les composés timbre-poste. Lyon, Presses universitaires de Lyon. • Baroni Marco, Guevara Emiliano, Pirrelli Vito (2009), Sulla tipologia dei composti N+N in italiano: přincipi categoriali ed evidenza distribuzionale a confronto. In: Ruben Benatti, Giacomo Ferrari and Monica Mosca (eds.), Linguistica e modelli tecnologici di ricerca (Atti del 40esimo Congresso della Società di Linguistica Italiana). Roma: Bulzoni, pp. 73-95. • Frantext 1 (2018) – base textuelle. http://www.frantext.fr/ • FrWac - Baroni, M. et al. 2009. The WaCky wide web: a collection of very large linguistically processed web-crawled corpora. Language Resources and Evaluation, vol. 43, no. 3, 209-226. • Grandi Nicola (2009), WhenMorphology 'Feeds' Syntax: Remarks on Noun > Adjectiveconversion in ItalianAppositiveCompounds. In: Montermini F., Boyé G., Tseng J. (eds.), SelectedProceedingsofthe 6th Décembrettes. Somerville, MA: CascadillaProceedings Project. • Koga Kentaro (2018), Un espace papeterie n’est-il pas une (simple) papeterie ?: composition binominale souscatégorisante. SHS Web Conf. , Vol. 46, 08004, Congrès Mondial de Linguistique Française - CMLF 2018. • Radimský Jan (2015), Noun+Noun Compounds in Italian. A corpus-based study.ČeskéBudějovice, University of South Bohemia. • Radimský Jan (2019), Les composés N-N de subordination : un paradigme émergent. Studia Romanica Posnaniensia 46/1 (2019), Adam Mickiewicz University Press, pp. 167-180. • Rainer Franz & Buridant Claude (2015), From Old French to Modern French. In: Muller Peter O. (ed.), Word-formation: an international handbook of the languages of Europe. Vol. 3. Berlin/Boston, Walter de Gruyter, pp. 1975-2000. • Traugott, ElizabethC., & Trousdale, Graeme (2013), Constructionalization and constructionalchanges. Oxford, Oxford University Press,