Taylor Berg-Kirkpatrick

Taylor Berg-Kirkpatrick is an Associate Professor in the Department of Computer Science and Engineering at the University of California, San Diego. He directs the BergLab research group, which focuses on developing machine learning methods for analyzing human language and artifacts, including historical documents, music, and multimodal data. His work bridges natural language processing, computational linguistics, and creative AI applications.

Research Interests: Berg-Kirkpatrick's research spans unsupervised learning for language understanding, structured prediction models, and computational analysis of human artifacts. The BergLab specializes in:

• Deciphering historical documents and languages using computer vision and NLP
• Developing generative models for music and audio synthesis
• Creating novel methods for low-resource and unsupervised machine learning
• Cross-modal analysis between text, audio, and visual data

Recent Publication Trends: His 15 most recent publications (2023-2024) demonstrate three key research vectors: advanced music generation using diffusion models and symbolic representations; computational analysis of historical artifacts including cuneiform tablets and early modern books; and foundational NLP work on model architectures, decoding methods, and multimodal systems. The work shows consistent innovation in generative AI, historical document processing, and efficient machine learning techniques.

Awards and Honors:

• NSF CAREER Award for 'Modeling Language Evolution via Deep Probabilistic Factorization' (2022)

Advising and Funding: Berg-Kirkpatrick has advised over 10 PhD students to completion on topics spanning NLP, music generation, and historical document analysis. His research is supported by the NSF CAREER grant and focuses on developing unsupervised methods for language evolution modeling. The BergLab maintains several open-source projects including Klavier (piano transcription), Ocular (historical document recognition), and Puck (high-speed parser).

Lab Direction: The BergLab at UCSD develops computational methods for analyzing diverse human artifacts through interdisciplinary collaborations with historians, musicians, and archaeologists. Current projects include computational bibliography of early modern books, decipherment of historical ciphers, and generative models for symbolic music.