What I do

I am a theoretical and computational linguist: I build formal models of human grammar based on empirical data from natural languages.

In my research, I develop a mathematical characterization of syntactic dependencies such as case assignment. By linking linguistic theory with results from formal language theory, I derive properties of syntactic parsing, which can inform general cognition as well as computer linguistics. I collect data by doing fieldwork: I've worked with speakers of Turkic languages (Karakalpak, Kyrgyz, Turkish) and also with speakers of Nepali and Yaadre.

I also build software for natural language processing for projects in digital humanities. Then, I work with statistical models to solve problems such as named entity recognition (NER) and text classification.

I work as data analyst for the Austrian Centre for Digital Humanities and Cultural Heritages (ACDH-CH) at the Austrian Academy of Sciences while I wrap up my Ph.D. studies in linguistics at UConn. Previously, I've spent a semester as visiting student at Stony Brook University, and studied linguistics and computer science in Vienna, Austria.

Download my full CV in English or auf Deutsch (last updated: February 2022)


Email me for copies of non-linked attachments. I'm always looking forward to receiving comments. Please don't hesitate to reach out!

Accusative case in Turkish action nominals and some light-verb constructions
  • Conference presentation at Turkologentag 2023, University of Vienna
Object shift in American Sign Language (ASL) and Brazilian Sign Language (Libras)
(with Ronice Müller de Quadros, Emily Jo Noschese, and Diane Lillo-Martin)
ASL and Libras have an object-shift construction by which the canonical SVO order is changed to SOV. In both sign languages, this ordering is mandatory for V marked with durative/continuative aspect (reduplicated movement), optional for V that agrees with O in locus, and not allowed with plain V. When V agrees with O in handshape, ASL requires OV ordering whereas Libras allows both OV and VO ordering. We present an analysis that derives these data with a combination of syntactic movement of O and violable, equally-ranked PF-constraints as proposed by Bobaljik & Wurmbrand (2012). Unlike Matsuoka's (1997) and Braze's (2004) pro- posals, we do not move V to a head on the right in violation of the Final-over-final constraint (Biberauer et al. 2014; Sheehan et al. 2017).
Germanic doubly-filled COMP
We derive doubly-filled COMP constructions in Bavarian and other Germanic varieties for embedded questions, successive-cyclic wh-movement, headed relative clauses, and head-less relative clauses by analyzing wh-words as basically feature-less (no Person, Gender, Number, Case, Animacy; only Wh/Rel). Together with Roberts's (2010) analysis of head-movement, this predicts correctly when wh-words and relativizers apper in Spec,CP and when in C.
  • ÖLT 45 at Universität Salzburg — Dec. 6–7, 2019.
  • Short manuscript in honor of Hamida Demirdache's 60th birthday; Feb. 12, 2021.
  • Olinco 5 at Palacký University Olomouc; June 10–12, 2021.
    • Proceedings paper: in print
  • CGG 30 at Oviedo — June 30–July 2, 2021.
An implicational universal linking c-command and movement
(with Thomas Graf)
Based on insights from mathematical linguistics, we argue that syntactic dependencies spanning independent subtrees, that is: not expressible via c-command relations, must be established by movement.
A formal account of Dependent Case Theory
I formalize DCT into Minimalist Grammars (Stabler 1997) with MSO-constraints on derivation trees (Graf 2013), and I argue that a formalization based on licensing by covert movement is not possible as it increases the complexity class of natural language. I provide a formalism that allows to circumvent this problem (to suspend the SMC) without incurring this blow-up in complexity.
Turkic nominalized clauses
I investigate the possibility of raising-to-possessor in Kyrgyz participial relative clauses with genitive-marked subjects. (The WAFL paper argues against it, but in light of new data, the TU+ paper argues in favor of it.)
  • ConCALL-3 at Indiana University, Bloomington, IN (talk) — March 2–4, 2018.
  • WAFL 14 at MIT, Cambridge, MA (talk) — October 19–21, 2018.
  • TU+ 4 at NYU, New York, NY (talk) — February 16–17, 2019.
    • Handout
I analyze clause-type triggered differential subject marking in Turkish by arguing that factive nominalized argument clauses are complex NPs with a silent head noun.
  • TU+2 at Indiana University, Bloomington, IN (poster, cancelled) — Nov 19–20, 2016.
  • ConSOLE 25 in Leipzig, Germany (talk) — Jan 4–6, 2017.
German impersonal passives
German has passives that retain accusative case and must receive a generic interpretation. I argue that they are impersonal actives.
  • Manuscript, May 2017.
The real(is) distinction in before- and after-clauses.
(with Marcin Dadan, Kadir Gökgöz, Jayeon Park, Yongsuk Yoo)
We compare before- and after-clauses in German, French, Japanese, Polish, Korean, and Turkish. We provide a uniform analysis for the different morphological strategies these languages use to mark before clauses. We argue that the intra-language differences between before and after are due to different selectional properties for mood, which stems from realis vs. irrealis presuppositions.
  • GLOW 39 in Göttingen, Germany (poster) — Apr 5–7, 2016.
  • FASL 25 at Cornell University, NY (talk) — May 13–15, 2016.
Turkish polar question particle
Based on evidence from conditionals I argue that the question particle "mI" triggers movement of the focused constituent.
  • TU+1 at UMass Amherst (talk) — Nov 21–22, 2015.
    • Proceedings paper: draft
    • Proceedings paper available from Amazon
The German relative pronoun "welch-" is an agreeing complementizer.
  • CGG 24 in Madrid, Spain (talk) — May 28–29, 2014.
  • RALFe at Paris 8, France (talk) — Oct 9–10, 2014.
    • Handout
Turkish accusative as topic marker.
  • ÖSKL 6 in Vienna, Austria (talk) — Nov 30–Dec 1, 2013.


The best way to reach me is by email.

You can write to me in English or German or French. I can read Turkish and SerBoCroatian, but I'll struggle to compose a response.

My preferred pronoun is her.