Description

One major focus of my research is the use of large language models to support endangered languages. I developed the first translator for Owens Valley Paiute by proposing a new translation paradigm: LLM-assisted rule-based machine translation (LLM-RBMT). We show that this approach performs well even without parallel corpora (a common limitation for low-resource languages). This hybrid approach combines LLMs and lexicons with expert-defined translation rules, enabling functional machine translation where traditional methods fail.

I also mentor students working on RAG-augmented chatbots that can interact with external sources like grammars and ethnographies to answer linguistic questions. Our public tools include a sentence builder, an English to OVP translator, and a digital dictionary to support both linguistic exploration and community use. This work sits at the intersection of AI and linguistic justice: our goal is to build tools that serve Indigenous communities.

People

  • Bhaskar Krishnamachari - University of Southern California - Professor
  • Diego Cuadros - Loyola Marymount University - Undergraduate Student - Computer Science
  • Jared Coleman - Loyola Marymount University - Assistant Professor of Computer Science
  • Kira Toal - Booz Allen Hamilton - Software Engineer
  • Nick Leeds - Loyola Marymount University - Undergraduate Student - Computer Science

Publications

  1. Community Voices on Digital Tools for Owens Valley Paiute - LMU Digital Commons - 01/01/2025
  2. No-Resource Languages - AmericasNLP @ NAACL 2024 - 4th Workshop on NLP for Indigenous Languages of the Americas - 01/01/2024

Media