Research
Bridging field linguistics and computational methods to preserve endangered languages and advance low-resource NLP.
Bulang Language Field Research
15 hrs/weekInitiated and conducted two field research expeditions to document the unwritten Bulang language, an endangered language with no standardized writing system. Established comprehensive linguistic database through systematic data collection.
Key Contributions
- Conducted two field trips to Bangbing village in Yunnan province
- Recorded 1,000+ lexical items in IPA (International Phonetic Alphabet)
- Analyzed phonological inventory: identified 50+ consonants and 30+ vowels including diphthongs
- Organized recordings into structured Bulang linguistic database
- Created comprehensive documentation with audio recordings and phonetic transcriptions
Bulang Checked Syllables & Pitch-Tone Research
10 hrs/weekInvestigating the relationship between pitch patterns and checked endings (glottal stops or consonant finals) in Bulang dialects (Avar and Bulangshan). Testing hypothesis that checked endings are determined by pitch rather than tone, with implications for the broader Tibeto-Burman language family.
Key Contributions
- Conducted comparative phonological analysis across Avar and Bulangshan dialects
- Testing novel theory: checked endings determined by pitch rather than traditional tone features
- Maintained weekly email updates and regular Zoom meetings with Prof. Chambless
- Planned milestone: Upload Bulang corpus to Duke research data repository
- Preparing findings for presentation at Duke University research seminar
Naxi Dongba Script Translator
15 hrs/weekDesigned a novel multi-agent AI pipeline for translating the endangered Dongba script, one of the few pictographic writing systems still in use. The system addresses unique challenges in low-resource NLP through innovative decomposition techniques.
Key Contributions
- Developed multi-agent architecture for sentence deconstruction and semantic analysis
- Implemented computer vision models for pictographic character recognition
- Advanced methodologies for low-resource language processing
- Integrated linguistic knowledge with deep learning approaches
Chinese Sign Language Recognition & Assistive Technology
5 hrs/weekConducted needs assessment and developed assistive AR technology for the deaf community. Combined fieldwork interviews with technical implementation to create user-centered solutions.
Key Contributions
- Collected and processed video dataset of Standard Chinese Sign Language
- Conducted in-depth interviews with deaf students to identify real-world challenges
- Trained machine learning models for sign language recognition
- Designed AR-based assistive technology prototypes