[PLing] CfP WS at SLE2017: What is in a morpheme? Theoretical, experimental and computational approaches to the relation of meaning and form in morphology

Stela Manova stela.manova at univie.ac.at
Tue Oct 11 12:38:55 CEST 2016

What is in a morpheme? Theoretical, experimental and computational approaches to the relation of meaning and form in morphology 

Dates: 10 – 13 September 2017 (50th SLE Meeting)
Location: Zurich, Switzerland
Call Deadline: 11 November 2016

Organizers: Stela Manova1, Harald Hammarström2 and Itamar Kastner3
                     1University of Vienna (Vienna), stela.manova (at) univie.ac.at <http://univie.ac.at/>
                     2Max Planck Institute for the Science of Human History (Jena), hammarstroem (at) shh.mpg.de <http://shh.mpg.de/>
                     3Humboldt University (Berlin), itamar.kastner (at) hu-berlin.de <http://hu-berlin.de/>

There are enough examples in science that obvious things are the most difficult to explain: issues such as how inorganic matter turns into organic or how a child learns to understand language. There is a similar problem in morphology. It is well known that morphemes consist of phonemes but only the former can be associated with meaning (systematically); and it is a non-trivial question how exactly this association happens.    
There are three possible ways to approach the relation of meaning and form:
A.  Form and meaning emerge simultaneously;
B. The association is from meaning to form;
C. The association is from form to meaning. 

The most important difference between these scenarios consists in the fact that in scenarios B and C meaning may be assigned at the level of word, i.e. one may claim that morphemes do not have meaning of their own or even that there are no morphemes at all (in scenario B). (Information (syntactic/morphological/morphosyntactic) that does not refer to (phonological) form is called ‘meaning’ in this proposal.)

Theoretical, experimental and computational linguistics approach word structure from different perspectives and seem to diverge with respect to which is the “right” scenario. Theoretical linguistics is interested in generalizations over meaning (features) (scenarios A and B), both within languages and typologically: e.g., only a language with plural can have dual; or no language makes more gender distinctions in the non-singular than in the singular (Greenberg 1963). Experimental linguistics researches perception, parsing, processing and production of word structure; computational linguistics is focused on parsing and distribution of word structure. Consequently, both experimental and computational linguistics follow scenario C and their findings seem to contradict theoretical linguistics (see below). Nevertheless, theoretical linguists (seem to) agree that speakers have somewhat reliable intuitions about n-gram frequency over sub-word units. Thus, the goals of this workshop are threefold: to encourage interdisciplinary discussion; to clarify and unify assumptions; and to pave the way for collaboration.

Let us first illustrate the different scenarios. Minimalist Morphology (MM) (Wunderlich 1996) is an example of scenario A. In MM, a morpheme has form and meaning; (inflectional) morphemes are heads; a(n) (inflectional) morpheme minimally includes a representation of its phonological form, a specification of the base’s category, and an output specification. (1) gives the specification of the 2 Sg marker -st in German (Stiebels 2011):

(1)                    /st/; [+min]; [+2]/+V    
[+min] indicates that the form is bound, i.e. a morpheme; [+2] means that it contributes the specification of 2 person and  “+V” indicates that -st attaches to verbs. The slash / stands for “output/input”.

In Realizational Morphology (RM), theories such as Paradigm Function Morphology (PFM) (Stump 2001) and Distributed Morphology (DM) (Halle & Marantz 1993), meaning and form are modeled separately and semantic derivation precedes formal derivation, the so-called late insertion (scenario B). Roughly, one can predict form based on meaning, while the opposite does not hold and therefore the form-to-meaning direction is not activated in RM.

PFM manipulates morphosyntactic property sets (2). A PFM representation from Stewart & Stump (2007):

(2)             PF(⟨L, σ⟩) = ⟨R, σ⟩

The value of the paradigm function (PF) of a paradigm cell ⟨L, σ⟩ (L stands for ‘lexeme') is the pairing of this cell’s realization R with the morphosyntactic property set σ. Such a theory does not necessarily need morphemes, although in the different versions of PFM one finds ‘roots' and ‘stems'.

DM relies on syntactic structures and ‘morpheme’, [past] in (3) below, is an abstract unit that refers to a syntactic terminal node (INFL in this case) and its content, not to the phonological expression of that terminal. A DM representation from Bobaljik (2015):

(3)     Vocabulary of English (fragment)
          a. [past] ↔ -t /]V __ ; where V ∈ {dream, dwell etc.}
          b. [past] ↔ Ø /]V__ ; where V ∈ {run, hit, fly etc.}
          c. [past] ↔ -d /] V__

To explain the fact that in DM syntactic structure derives morphological structure, Müller (2016) refers to the meaning-form dichotomy as two different dimensions of a linguistic unit: a representational and an algorithmic one respectively.

In the RM literature, the assumption that semantics precedes exponence is claimed to make RM superior in comparison to incremental theories of morphology that follow scenario A because in RM one manipulates the semantic side of the derivation, which takes place at an abstract level and is always compositional, while exponence (formal derivation) often entails idiosyncrasies.

On the other hand, affixes are directly accessible through their form, i.e. affixes can be identified and processed even without having a contentful stem to attach to, as evidenced by recent psycholinguistic studies, which speaks in favor of scenario C. For example, Crepaldi et al. (2016) uses as primes non-words and demonstrates that they facilitate lexical decisions to target words ending with the same suffix as well as that the priming effect depends on the affix position in the non-word, i.e. a prefix in the non-word does not facilitate the recognition of a suffix in the target word, even if the prefix and the suffix share the same form. Lázaro et al. (2016) uses suffixes as primes and shows that the prime suffix facilitates the recognition of words ending with that suffix. Both studies conclude that the priming effect of suffixes is not orthographic but morphological, i.e. the effect holds for derived and pseudo-derived words (such as corn:er) but was not found for simplex words as targets. Findings similar to those in Crepaldi et al. (2016) and Lázaro et al. (2016) are also reported in Beyersmann et al. (2016).  Additionally, Manova & Brzoza (2015) shows that native speakers, if provided with a list of existing and non-existing suffix combinations without bases (i.e. without semantic cues), can, with a remarkably high accuracy, judge which combinations exist and which do not.

In computational linguistics, the “emergence” of morphological structure invariably starts from form, since corpora are not annotated for meaning (Baroni 2003, among others). An approach called Unsupervised Learning of Morphology (ULM) takes raw data such as an unannotated corpus and provides a list of affixes and stems that occur in those data. The most popular strategies on which ULM relies include comparison, grouping and weighting of substrings (of letters), see Hammarström and Borin (2011) for an overview of ULM research. Only after the extraction of potentially related forms, semantics (in terms of semantic similarity) can be assigned (Baroni et al. 2002).

The workshop will provide a platform for exchange of ideas and for an interdisciplinary discussion of the meaning-form issue in morphology. It will bring together theoretical and computational linguists (including computer scientists), psycho- and neurolinguists (including psychologists), fieldworkers and typologists. The questions to be addressed include, but are not limited to, the following:

1.     What information is encoded in a morpheme?
2.     Does an analysis with emphasis on either meaning (scenario B) or form (scenario C) provide evidence for a (complete) separation of form and meaning in the morpheme?
3.     Could it be that a morpheme relates meaning and form and semantic stimuli activate derivation through meaning, while formal stimuli activate access through form?
4.     How does morphology “emerge” in fieldwork, i.e. how does a fieldworker decide that something is a morpheme, is it according to A, B or C?
5.     How does morphology “emerge” in child language?
6.     What exactly does a language borrow when it borrows morphological structure such as, e.g., a plural nominal marker, if that language already has ‘plural' and its speakers are not expected to be able to perform a morpheme analysis of 
	the donor language’s words?
7.     If important generalizations are (necessarily) stated over either meaning or form, how are the two types of generalizations related to one another; and are they both needed for an adequate characterization of speakers' knowledge of their 
8.     Can a computational analysis based on n-gram frequency distributions and distributional semantics account for the kinds of generalizations that interest theoretical linguists and motivate the B (or A) perspective?

As an alternative, non-linguistic source of inspiration, we would like to turn your attention to the following video on how computers learn to understand pictures: https://www.youtube.com/watch?v=40riCqvRoMs <https://www.youtube.com/watch?v=40riCqvRoMs> (the speaker, Fei-Fei Li, is an Associate Professor of Computer Science at Stanford University). Computer vision is one of the most important areas of research in machine learning and many striking analogies with linguistic analyses can be made.

We invite abstracts that are no longer than 300 words, excluding examples and references, and tackle any aspect of the form-meaning issue in morphology. Papers that report recent psycho-, neuro-, computational and theoretical linguistics research are particularly welcome. Please submit a pdf of your abstract to stela.manova at univie.ac.at <mailto:stela.manova at univie.ac.at> by November 11, 2016.

Important Dates
11 November 2016: deadline for submission of 300-word abstracts to the workshop organizers
20 November 2016: notification of acceptance by the workshop organizers
25 November 2016: submission of the workshop proposal to SLE
25 December 2016: notification of acceptance of workshop proposals by SLE
15 January 2017: deadline for submission of full abstracts to SLE for review
31 March 2017: notification of paper acceptance
10-13 September 2017: SLE conference in Zürich

Baroni, M. (2003). Distribution-driven morpheme discovery: A computational/experimental study. In G. Booij and J. van Marle (eds.), Yearbook of Morphology 2003, 213-248. Dordrecht: Springer.
Baroni, M.; J. Matiasek and H. Trost. (2002). Unsupervised discovery of morphologically related words based on orthographic and semantic similarity. In Mike Maxwell (ed.), Proceedings of the Workshop on Morphological and Phonological Learning of  	  	 
	ACL/SIGPHON-2002, 48-57. East Stroudsburg PA: ACL.
Beyersmann, E; JC Ziegler; A Castles, M Coltheart, Y Kezilas & J Grainger (2016). Morpho- orthographic segmentation without semantics. Psychon Bull Rev 23(2): 533-9. Abstract available at: https://www.ncbi.nlm.nih.gov/pubmed/26289649 <https://www.ncbi.nlm.nih.gov/pubmed/26289649> 
Bobaljik, J.  (2015). Distributed Morphology. Ms. University of Connecticut. Available at: http://bobaljik.uconn.edu/papers/DM_ORE.pdf <http://bobaljik.uconn.edu/papers/DM_ORE.pdf>
Crepaldi D., L. Hemsworth, C. J. Davis & K. Rastle. (2016). Masked suffix priming and morpheme positional constraints. Q J Exp Psychol (Hove) 69(1): 113-28. Abstract at: https://www.ncbi.nlm.nih.gov/pubmed/25760942 <https://www.ncbi.nlm.nih.gov/pubmed/25760942> 
Greenberg, J. H. (ed.) (1963). Universals of Human Language. Cambridge, Mass: MIT Press.
Halle, M. & A. Marantz (1993). Distributed morphology and the pieces of inflection. In Hale K. and S. J. Keyser (eds.), The view from building 20, 111-176. Cambridge, MA: MIT Press.
Hammarström, H. & L. Borin (2011). Unsupervised Learning of Morphology. Computational Linguistics 37(2): 309-350.
Lázaro, M, V. Illera & J. Sainz (2016). The suffix priming effect: Further evidence for an early morpho-orthographic segmentation process independent of its semantic content. Q J Exp Psychol (Hove) 69(1): 197-208. Abstract at:  
	https://www.ncbi.nlm.nih.gov/pubmed/25801451 <https://www.ncbi.nlm.nih.gov/pubmed/25801451> 
Manova, S. & B. Brzoza. (2015). Suffix combinability and the organization of the mental lexicon. Paper presented at the 10th Mediterranean Morphology Meeting (MMM), Haifa, Israel, 7-10 September 2015. Handout available at: 
	http://homepage.univie.ac.at/stela.manova/uploads/1/2/2/4/12243901/manova_brzoza_mmm2015.pdf <http://homepage.univie.ac.at/stela.manova/uploads/1/2/2/4/12243901/manova_brzoza_mmm2015.pdf> 
Müller, G. (2016). Minimize Satisfaction in Harmonic Serialism. Paper presented at the workshop The Word and the Morpheme. Humboldt University zu Berlin, 22.9-24.9.2016. Abstract available at: https://www.angl.hu- <https://www.angl.hu-/>
	berlin.de/department/staff/artemis_alexiadou/abstracts/muller.pdf <http://berlin.de/department/staff/artemis_alexiadou/abstracts/muller.pdf> 
Stewart, T. & G. Stump (2007). Paradigm Function Morphology and the Morphology–Syntax Interface. In G. Ramchand & C. Reiss (eds.), The Oxford handbook of linguistic interfaces, 383-421. Oxford: Oxford University Press.
Stiebels, B. (2011). Minimalist Morphology. Ms. University of Leipzig. Available at: http://home.uni-leipzig.de/stiebels/papers/handbook_morphology_stiebels_mm-2011.pdf <http://home.uni-leipzig.de/stiebels/papers/handbook_morphology_stiebels_mm-2011.pdf> 
Stump, G. T. (2001). Inflectional morphology. Cambridge: Cambridge University Press.
Wunderlich, D. (1996). Minimalist morphology: the role of paradigms. In G. Booij & J. van Marle (eds.),Yearbook of Morphology 1995, 93–114. Dordrecht: Kluwer.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.univie.ac.at/pipermail/pling/attachments/20161011/a205fe9f/attachment-0002.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CfPSLE2017Workshop.pdf
Type: application/pdf
Size: 150986 bytes
Desc: not available
URL: <https://lists.univie.ac.at/pipermail/pling/attachments/20161011/a205fe9f/attachment-0001.pdf>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.univie.ac.at/pipermail/pling/attachments/20161011/a205fe9f/attachment-0003.html>

More information about the PLing mailing list