computational modeling

Welcome Dunstan Brown

The morphological systems group is excited to be hosting Dunstan Brown, from University of York, for a two month visit to OSU. He is here to work with Andrea on a project modeling Greek nominal stress in DATR, and to talk all things defectiveness. (Check out his and Neil Bermel‘s “Feast and Famine” grant project on defectiveness and overbaundance!) And it is a great opportunity for our local community of morphologists to talk with Dunstan about his/their research: morphology and its interface to syntax, computational linguistics, Slavic languages, Network Morphology, Canonical Typology, inflectional complexity — so many points of shared interest!

Dunstan will be giving a colloquium talk in the department on October 29.

Welcome, Dunstan!

Morphological Systems Group organizes AIMM5

 

Screenshot of AIMM5 GatherTown spaceScreenshot of AIMM5 poster session in GatherTown

We got to welcome more than 160 morphologists from around the world to OSU (virtually) for the 5th American International Morphology Meeting (AIMM5) last weekend. It was four busy but exciting days of stimulating talks and interesting discussion. (Check out this picture of one of the poster sessions that took place in Gather.Town!) We hope that everyone enjoyed the conference. We certainly did!

It was also an opportunity to show off some of the morphological research happening here at OSU. The program included five presentations from our group:

  • Martha Booker Johnson and Andrea D. Sims, “Using word vectors to investigate semantic transparency cross-linguistically”
  • Kyle Maycock and Andrea D. Sims, “Albanian second-position clitics as edge inflection: Evidence from cumulative exponence in the noun phrase”
  • Connor Rouillier, “The effect of event structure on subject-verb agreement in Najdi Arabic”
  • Noah Diewald, “Wao Terero lexical suffixes: Realization at the lexical semantic-discourse interface”
  • Micha Elsner and Andrea D. Sims, “Analogical modeling of morphology for L1 effects in language contact”

There were also presentations from OSU “friends of the morphology lab” Brian Joseph, Shuan Karim, and John Grinstead (with colleagues).

We look forward to AIMM6, to be held at the University of California, San Diego in 2023!

Grace LeFevre defends B.A. thesis

Congratulations to Grace LeFevre, who successfully defended her honors B.A. thesis, Quantifying Paradigm Shape in Spanish Verbs! The thesis was co-advised by Micha Elsner and Andrea Sims. A paper based on the thesis has already been published in the 4th Proceedings of the Society for Computation in Linguistics.

Abstract: This thesis computationally models “paradigm shape,” a type of morphological structure that I define by the implicative relations holding among the forms in an inflectional system. Since implicative structure binds the forms in an inflectional system together (Wurzel, 1989), paradigm shape reflects the predictable ways that allomorphs occur in parallel paradigm cells across inflection classes in some languages. Maiden (2005)’s analysis of how certain Romance verbs changed over time in order to conform to existing paradigm shapes highlights the significance of this structure as a historical and cognitive organizing principle. However, paradigm shape has not been computationally formalized in a gradient or replicable way. Using information-theoretic entropy as defined by Shannon (1948), I develop a method to quantify paradigm shape and I apply it to Spanish verbs as a test case. The method bridges the gap between formal work on the organization of the stem space (e.g. Maiden, 2005; Boye and Cabredo Hofherr, 2006) and computational work on quantifying predictability in inflectional systems (e.g. Ackerman and Malouf, 2013; Stump and Finkel, 2015). In doing so, it jointly models the distributions of stems and affixes to compute sets of values that characterize the shapes of Spanish verb classes. Comparison of these values across classes captures partial parallelism between them, enabling identification of both allomorphic and distributional class structures (Baerman et al., 2017). These results with Spanish verbs highlight that my method provides a computational means of capturing multiple aspects of inflection class structure in a way that is replicable and extendable to other languages. Potential directions for future work include testing the limits of the method’s usefulness on known morphologically difficult systems and applying the method to other Romance languages at various stages of historical development.

Society for Computation in Linguistics paper

Grace LeFevre, Micha Elsner and Andrea Sims had their paper “Formalizing Inflectional Paradigm Shape with Information Theory” published in the Proceedings of the Society for Computation in Linguistics, vol. 4. The paper is based on Grace’s B.A. thesis work. She did really impressive work and we are happy to see it make it out into the world!

Abstract: “Paradigm shape,” our term for the morphological structure formed by implicative relations between inflected forms, has not beenformally quantified in a gradient manner. We develop a method to formalize paradigm shape by modeling the joint effect of stem alternations and affixes. Applied to Spanish verbs,our model successfully captures aspects of both allomorphic and distributional classes.These results are replicable and extendable to other languages.

Joint morphology and computational linguistics seminar

This semester Andrea Sims and Micha Elsner are holding joint meetings of Andrea’s graduate morphology seminar and Micha’s graduate computational linguistics seminar. The joint seminar, focusing on Models of Morphological Learning and Change, is designed to bring together students with backgrounds in morphology, computational linguists, language acquisition, and historical and sociolinguistics… but not necessarily more than one of these. It is an experiment in talking across subdisciplinary boundaries, with the hope that the whole will be more than the sum of its parts (not unlike morphological structure!). We are excited to see what projects will develop!

Seminar description: Where do languages come from, and how do they evolve? We learn the languages we speak as infants or students; as adults, we transmit them to new generations of speakers. In a variety of linguistic sub-areas, researchers have claimed that this process of iterated language learning influences the kinds of languages which exist in the world (language typology) and the process of language change over time. Many of these researchers have proposed computational models of this process, enabling the rapid simulation of “learners” exposed to different language inputs, and of many generations of “teaching” and “learning”.

This seminar will investigate the learning process, with special reference to the case of inflectional morphology (grammatical forms of a word, such as singular cat ~ plural cats). We will bring together research in several areas of linguistics in order to discover how the different perspectives taken across sub-communities combine (or fail to combine!) to address the problem. The reading list will cover:

  1. Typology of inflectional systems: what sort of languages are out there
  2. Learning-based theories of morphological typology and change
  3. Cognitive models of morphological learning
  4. Engineering models of inflection prediction
  5. Computational work on iterated language learning

Texas Linguistics Society

In September, Andrea Sims gave a keynote talk, “Morphological connectivity in the mental lexicon,” at the 17th Texas Linguistics Society meeting, in Austin, TX.

Check out the slides of the talk: sims-TLS2017-slides.

Thanks to all of the organizers for the invitation and a great conference!