Researcher

Brink van der Merwe

Brink van der Merwe is a computer scientist at Stellenbosch University whose research spans natural language processing, automata theory and regular-expression security. He earned a PhD at Texas A&M University in 1995, coauthored a grammatical error correction model in 2013 and studied exponential-time behavior in Java regex engines in 2014. In 2023, he formalized byte pair encoding tokenization for language models and NLP systems; in 2024, he coauthored an algorithm for finite automata over BPE tokenizations. He led Computer Science at Stellenbosch in 2025 and linked entrepreneurial teaching and close industry ties with alumni success. As of October 2026, he is a professor and head of the Computer Science Division, with interests in machine learning for NLP and regular-expression denial-of-service attacks.

Timeline

  1. 1995

    PhD at Texas A&M

    He earns a PhD at Texas A&M University in 1995.

  2. 2013

    A grammar correction model

    In 2013, he and Jan Buys publish a tree-transducer model for grammatical error correction.

  3. 2014

    Modeling Java regex backtracking

    With Martin Berglund and Frank Drewes, he develops a formal model of Java regex matching and static analyses for expressions that can trigger exponential-time behavior in 2014.

  4. 2023

    Formalizing BPE tokenization

    His 2023 paper formalizes byte pair encoding tokenization as used in large language models and other NLP systems. He serves on the organizing committee for FSMNLP 2023.

  5. 2024

    A DFA for BPE tokenization

    With Martin Berglund and Willeke Martens, he presents an algorithm in 2024 for building deterministic finite automata that operate directly on byte pair encoding tokenizations.

  6. 2025

    Head of Computer Science

    In 2025, he heads Computer Science at Stellenbosch and argues that entrepreneur-led teaching and close industry ties contribute to alumni success.

  7. 2026

    Regex and automata algorithms

    In June 2026, he and his coauthors introduce a minimum-feedback-node strategy for selective memoization in backtracking regex matching. In July 2026, he coauthors an analysis of acceptance, emptiness and equivalence for symmetric-difference automata.

  8. As of October 2026

    Professor and division head

    As of October 2026, he is a professor and head of division in Stellenbosch University's Computer Science Division. His research interests include machine learning for NLP and regular-expression denial-of-service attacks.

Career

  1. NowStellenbosch UniversityProfessor and Head of Division

Education

  1. 1995Texas A&M UniversityPhD

Sources and credits

Photo: We have no published photo of Brink van der Merwe yet. Send us a free-licensed photo.

  1. van der Merwe Brink jucs.org. Checked 20 October 2026.
  2. A Tree Transducer Model for Grammatical Error Correction - ACL Anthology aclanthology.org. Checked 20 October 2026.
  3. Analyzing Catastrophic Backtracking Behavior in Practical Regular Expression Matching arxiv.org. Checked 20 October 2026.
  4. Formalizing BPE Tokenization arxiv.org. Checked 20 October 2026.
  5. FSMNLP 2023 in Stellenbosch fsmnlp2023.github.io. Checked 20 October 2026.
  6. Constructing a BPE Tokenization DFA arxiv.org. Checked 20 October 2026.
  7. Why Stellenbosch keeps churning out successful start-ups techcentral.co.za. Checked 20 October 2026.
  8. Selective Memoization for Efficient Backtracking Regular Expression Matching arxiv.org. Checked 20 October 2026.
  9. Algorithms and fine-grained complexity for nondeterministic and symmetric difference automata arxiv.org. Checked 20 October 2026.
  10. Stellenbosch CS/RW cs.sun.ac.za. Checked 20 October 2026.
  11. Brink van der Merwe | CAIR cair.org.za. Checked 20 October 2026.