Publications

Last updated Feb 27, 2023 Edit Source

1. ▲ paper at a major NLP/CL conference or journal (e.g. *CL, LREC, TACL).
   
2. ▩ paper at a NLP/CL workshop or smaller conference (e.g. SIGMORPHON, Syntaxfest).
   
3. ◉ non-archival talk (e.g. SIGTYP abstract).
   
4. (none) preprint.

Brett Reynolds, Aryaman Arora, Nathan Schneider. CGELBank: CGEL as a framework for English syntax annotation.

▩ Jordan Kodner, …, Aryaman Arora, …, Ekaterina Vylomova. SIGMORPHON–UniMorph 2022 Shared Task 0: Generalization and Typologically Diverse Morphological Inflection. SIGMORPHON.

▩ Khuyagbaatar Batsuren, Gábor Bella, Aryaman Arora, …, Ekaterina Vylomova. The SIGMORPHON 2022 Shared Task on Morpheme Segmentation. SIGMORPHON.

▲ Aryaman Arora. Universal Dependencies for Punjabi. LREC.

• code

▲ Aryaman Arora, Nitin Venkateswaran, Nathan Schneider. MASALA: Modelling and analysing the semantics of adpositions in linguistic annotation of Hindi. LREC.

• code

▲ Khuyagbaatar Batsuren*, Omer Goldman*, …, Aryaman Arora, …, Ekaterina Vylomova. UniMorph 4.0: Universal Morphology. LREC.

▲ Aryaman Arora, Clara Isabel Meister, Ryan Cotterell. Estimating the entropy of linguistic distributions. ACL.

• code

▲ Aryaman Arora, Adam Farris, Samopriya Basu, Suresh Kolichala. Computational historical linguistics and language diversity in South Asia. ACL.

◉ Aryaman Arora, Nathan Schneider, Brett Reynolds. A CGEL-formalism English treebank. MASC-SLL, Philadelphia, USA (April 30, 2022).

• code

◉ Adam Farris*, Aryaman Arora*. DIPI: Dependency Parsing for Ashokan Prakrit Historical Dialectology. Towards a comparative historical dialectology: evidence from morphology and syntax @ DGfS, Tübingen, Germany (February 23–25, 2022).

• code

▩ Adam Farris*, Aryaman Arora*. For the purpose of curry: A UD Treebank for Ashokan Prakrit. Universal Dependencies Workshop.

• code

▩ Aryaman Arora, Adam Farris, Gopalakrishnan R, Samopriya Basu. Bhāṣācitra: Visualising the dialect geography of South Asia. Workshop on Computational Approaches to Historical Language Change.

• code

Aryaman Arora, Ahmed Etebari. Kholosi Dictionary.

Aryaman Arora, Nitin Venkateswaran, Nathan Schneider. Adposition and case supersenses v1.0: Guidelines for Hindi–Urdu.

▩ Aryaman Arora, Nitin Venkateswaran, Nathan Schneider. SNACS annotation of case markers and adpositions in Hindi. SCiL.

• code

▩ Michael Kranzlein, Emma Manning, Siyao Peng, Shira Wein, Aryaman Arora, Nathan Schneider. PASTRIE: A corpus of prepositions annotated with supsersense tags in Reddit International English. Linguistic Annotation Workshop.

• code

◉ Aryaman Arora, Nathan Schneider. SNACS annotation of case markers and adpositions in Hindi. SIGTYP.

• code

▲ Aryaman Arora, Luke Gessler, Nathan Schneider. Supervised grapheme-to-phoneme conversion of orthographic schwas in Hindi and Punjabi. ACL.

• code
• slides

• Story

  This was the first real project I worked on. I had no practical ML experience going in (beyond having watched a few lectures by Andrew Ng), had cold-emailed this professor, and it turned out his Ph.D. student Luke knew Hindi and had a project idea.

  The basic problem I had been dealing with for a while actually. In Hindi, the orthography has an inherent schwa attached after every consonant unless you use the vowel-killer (halant) or a conjunct consonant. However, in speech, this schwa is sometimes not pronounced; the most obvious instance of this is at the end of the word, e.g. orthographic ⟨ka-ra⟩ is pronounce /kar/. As an editor on Wiktionary I had been trying to write a rule-based convertor from Hindi orthography to pronunciation, and schwa deletion was a really annoying problem. Linguists don’t have a good story about schwa deletion; it is a prosodic phenomenon but also definitely interacts with morphology.

  So we just threw machine learning at the problem and got an insanely good score, schwa deletion is solved! But anyway, I do think about this problem occasionally still. In ~2021 I was thinking “why didn’t I use a biLSTM instead to have infinite context window in the prediction?” And in ~2023 I’m thinking “why didn’t I try harder to understand what algorithm was learned in the decision tree?”

Aryaman Arora, John R. McIntyre. Quasi-passive lower and upper extremity robotic exoskeleton for strengthening human locomotion. Sustainable Innovation.