Research Interests

Computational linguistics and NLP

My computational interests center around applying computational morphology and phonology to problems in natural language processing. While linguistic representations have long been employed in a variety of NLP tasks, these representations have been primarily morphological, syntactic, semantic, and discourse-analytic. My mission has been to expand the role of morphological, and especially phonological, representations in NLP. Some specific research directions include the following:

• Phonological representations for cross-lingual transfer learning
• Phonological representations (segments, features) in named entity recognition (NER)
• Predicting the forms of loanwords from phonological models
• Approximate phonological matching and entity linking
• The implementation and application of automatic (rule-based, unsupervised, and semi-supervised) syllabification
• Tools for compiling metarules (with phonology feature specifications) into finite state transducers

My interests in computational morphology are perhaps more straighforward:

• Developing more user-friendly and expressive frameworks for writing morphological analyzers, particularly for languages with non-concatenative morphology
• Rich morphological representations for machine translation from morphologically rich source langauges
• Machine translation into morphologically rich target languages

Finally, I have long-standing interests in computational historical phonology

• Test hypothetical proto-language reconstructions and sequences of sound changes through computational means
• Cognate prediction problem: given two corpora of related languages and a word from one corpus, predict the form of its cognate in the other language, even if it is not present in the corpus
Proto-language inference problem: given two corpora of related languages, infer a proto-language and two transducers (representing sequences of sound changes) such that the number of unique cognate pairs predicted from the proto-language and the two transducers is maximized

Theoretical and descriptive linguistics

I have long worked on languages of East and Southeast Asia. The specific languages and groups in which I am interested and which I have worked on are as follows:

• Hmong-Mien
• Western Hmongic
• Tibeto-Burman
• Tangkhulic
• Kuki-Chin
• Jingpho

Aside from specific languages, I am interested in a variety of linguistic subfields and issues. Here is a representative outline of my theoretical interests:

• Phonology
• Tone
• Phonation type/register
• Chain shifts and other counterfeeding opacity
• Abstractness of phonological relationships
• Phonetics-Phonology interface
• Phonology-Morphology interface
• Morphology
• Compounding
• Process morphology
• Reduplication
• Phonological constraints on morphotactics
• Affix ordering
• Historical Linguistics
• Comparative reconstruction
• Reconstructing phonological grammars
• Speaker misunderstanding and misinterpretation as a source of linguistic innovation
• Language contact
• Language Description