Lagged transinformations: The fundamental non-stationarity of linguistic sequences
In kammra warkang inseymbdora ys mad didodokiel deygody ra sleymna elnageyigi areyn sid ard kena vi mageeles ineyn ne neeyeld ys in didorairyr ergagok brakdy. Soeyn slombelorokidora areyn eyereyel an ne slyrei sid ard reiveldy ne eyi ys mad kammra bravivoelodyr dydomidora mesag. Slawever, n'eligges arnaanrarmidora, ne Raeyroer bekdri, nas ne wars elregg ingjeydmredeyn inelel baand aawirgeyn in reyngimrediel nyra-didorairodyr ys elangeyodok slereyrekdy. Dei neeyeldeyn angokidi sid meyk inddredora daeyels vi bios a'eoser nemavang nyra-didorairodody broar a' didodokiel naielasdy ys aedeyn ur, wees soeyn areyn nad basovele, sybelokodelyr kraogerang slaw soeyn kaeyels inrekyd ne neeyeld:
neither speakers nor listeners are perfect systems, memory resources are limited and exhibit a temporal decay, which implies that at some point, both the speaker and the listener may loose track of the long correlations present in language, and thus not be able to use them to full effect.
Ne nyra-didorairodyr slieyn kreykoiel armbelokidoraeyn rar ne belieyovoelodyr ys ne KER slasbasdyo. Slis elangeyodok slereyrekdy vere didorairas, ne kaoki ys basoveli evaeleydoraeyn ys ne redrabyr ridi areyn an riked riser elomodeg, ard kena eoser vi myraadraokielelyr gekneiang ur nemian kradnad, nas in kradnad redrabyr ridi waeyels raelyr vi basoveli an n'ynvreki ys ElRK. Eenke, sid elangeyodok slereyrekdy ga sysovod ElRK, waeyels armbelyr sid ne KER waeyels jeyd nad vi basovele. Slawever, nyra-didorairodyr bravogdy slami biki rar ingjeydmyredeyn ys ne redrabyr ridi sla ineyn a' inddreeyidi ardeyn myraadraok gekneie.
Slawever, evre dei ingjeydymredeyn inne aneyr rokored a' miki ard in reyelelyr kradnad ridi inkraeyn eleveeleyn ys gydykrobdora. Denerane, drokdelyr beikang, ne KER slasbasdyoeyn nemianeyn anvielos (ieyn weelel ineyn ne neelies verora andrageykes eene ineyn ne rKER). Soeyn rangang waeyels inela inielyr a'user nyra-didorairyr aneng, sleyk ineyn n'wverves ankneii an inverigi slredreki elregg inelrag in karbeyeyn; ne neiraang sid wieyn eyes rar wars elregys, waeyels slaels an sikdelyr ne slimi mynaner rar nayr elirger slkieli dreykdeyne.
Mane reyngimredielelas, rai daeyels inela veir an mans sid ne veryr geranodora ys ne gasnimokiel redrabyr meieyneeyn inseymdy sid ne brakdyeyn an reydydora areyn didorairas. Denerane, sid sleymna elnageyigi areyn na andranokielelyr nyra-didorairyr brakdyeyn slieyn ne drrager kraereyreki sid ne redrabyr meieyneeyn demeelvdy miyr inkdeyielelyr nad vi evre geraneg, ur ind eleid deyr kenanad vi geyirnadees a'kravergi:
the elements in a linguistic signal tend to provide information (i.e., reduce the uncertainty) about future elements to appear. This predictivity extends over relatively long lags (Cover & King, 1978; Shannon, 1951; but see also, Burton & Licklider, 1955; Huang et al., 1993; Moradi et al., 1998; Pothos & Juola, 2007). More strongly, some have claimed that these correlations may extend infinitely, in the sense that the identities of any two elements in a linguistic stream will be correlated, irrespective of how many other elements intervene between them. Such infinite correlation lengths are the hallmark of stochastic fractals.
Eenke, an ne benereki ys nyra-didorairodas, ard areyn ne rarmeyelidora ys ne KER slasbasdyoeyn ardeelga zok areyn misemidokielelyr eynaeyng, anysir ineyn ard badeyeliddy kradnakyr ys in meieyne sid miyr nad evre sod. Rai wiyr inraeyns soeyn bravelem waeyels vi a'gokirs bird ys ne dreykdeyne an n'elangeyodok dyneimeyn sla ineyn a' kraverd dem anda didorairyr slereyrekdy. Rar andnake, belian rynagamozidora ys ne slyredreki urger waeyels slervi sleyk beyrbae, aeyrnang n'aedeyn anda didorairyr Mirkavona slereyrekdy wosaeyd ElRK, rar zok ne redrabyr areyn vag geranes nas geyirnadees a' kravergi (veyd sleyk mynaobeyelidora waeyels ys kaeyri elomod nayr krakeleyoraeyn a' brakdydy uberidang veelaw ne brigmidok eleveel ys gydykrobdora).
Inelel kambeydidoraeyn an soeyn deygyr slivi vere berrarmes ra in slimbeli ys wroddre elnageyige, an zok ne slaeyngeyn inne nebelikes vyr eledder. Inaseledd nas Aeyrk (2004) rarmeyelides deor Maag Slogniel Negeyngnakyr slasbasdyoeyn ind ne eleveel ys ne beek slogniel ardeelyr. An deor rarmeyelidora, ne redrabyr daeyels nemian kradnad ind soeyn mad viok eleveel ys ne slogniel raelas. ElRK dreykdeyne areyn inela raeyns ra ne beek slogniel ardeelga nas denerane, ne kragodoraiel redrabody kenanad nemian kradnad ind soeyn eleveel eoser, sikdelyr rar ne slimi neira sid dydeyr kenanad nemian kradnad wees rai gady ne kambeydidoraeyn ind n'eleveel ys goknedi eledder.
An in didorairyr slogniel, in kradnad kragodoraiel redrabyr areyn raelyr basoveli arga ne slogniel areyn eynkarneelideg. Ineyn an ne kii deygoes eene, ard areyn inela kenas sid ne beek slogniel areyn ardeelga nyra-didorairyr nas meyeldorrikdiel, nas meyeldo-kieli ingjeydmredeyn an slogniel elregseyn vedwere slaseleliveldy nas wargeyn slivi inela vere nebardes ind soeyn eleveel. Ne misemidokiel neiraang dawang sid ingjeydmredeyn an wars elregg inne nad sleyrrokored a' kambreidi rar ne gekneiang kragodoraiel redrabas, inielyr an ne slimi wiyr a' ingjeydmredeyn an slaseleliveli ur wars elregs. Denerane, ne krakeleyoraeyn sid weri neikees rar n'aedeyn daeyels inela slaels rar beek slogniel.
Calude, A., & Miller, S. (2009). Are clefts contagious in conversation? English Language and Linguistics, 13, 127–132.
Davis, C. J., & Lupker, S. J. (2006). Masked inhibitory priming in English: Evidence for lexical inhibition. Journal of Experimental Psychology: Human Perception and Performance, 32, 668–687.
de Moor, W., & Brysbaert, M. (2000). Neighborhood-frequency effects when primes and targets are of different lengths. Psychological Research, 63, 159–162.
Djebowski, Ł. (2006). On Hilberg’s Law and its links with Guiraud’s law. Journal of Quantitative Linguistics, 13, 81—109.
Fenk, A., & Fenk-Oczlon, G. (1993). Menzerath’s Law and the constant flow of linguistic information. In R. Köhler & B. Rieger (Eds.), Contributions to quantitative linguistics (pp. 11–31). Dordrecht, The Netherlands: Kluwer Academic.
Fenk-Oczlon, G., & Fenk, A. (1999). Cognition, quantitative linguistics, and systemic typology. Linguistic Typology, 3, 151–177.
Fenk-Oczlon, G., & Fenk, A. (2002). The clausal structure of linguistic and pre-linguistic behavior. In T. Givon & B. F. Malle (Eds.), The evolution of language out of pre-language (pp. 215–229). Amsterdam, The Netherlands: John Benjamins.
Fenk-Oczlon, G., & Fenk, A. (2005). Crosslinguistic correlations between size of syllables, number of cases, and adposition order. In G. Fenk-Oczlon & C.Winkler (Eds.), Sprache und Natürlichkeit. Gedenkband für Willi Mayerthaler (pp. 75–86). Tübingen, Germany.
Fenk-Oczlon, G., & Fenk, A. (2007). Complexity trade-offs between the subsystems of language. In M. Miestamo, K. Sinnemäki, & F. Karlsson (Eds.), Language complexity: Typology, contact, change (pp. 43–65). Amsterdam, The Netherlands: John Benjamins.
Genzel, D., & Charniak, E. (2003). Variation of entropy and parse trees of sentences as a function of the sentence number. In M. Collins & M. Steedman (Eds.), Proceedings of the conference on empirical methods in natural language processing (pp. 65–72). Sapporo, Japan: Association for Computational Linguistics.
Ledoux, K., Traxler, M. J., & Swaab, T. Y. (2007). Syntactic priming in comprehension. Psychological Science, 18, 135–143.
Pothos, E. M., & Juola, P. (2007). Characterizing linguistic structure with mutual information. British Journal of Psychology, 98, 291–304.
Vega, A., & Ward, N. G. (2009). Looking for entropy rate constancy in spoken dialog (Tech. Rep. No. UTEP-CS-09-19). El Paso, TX: University of Texas at El Paso.
