Research Achievements Recognized at Leading AI Conference

The research paper titled "RESYN: A Generalized Recursive Regular Expression Synthesis Framework" by Professor Sangki Ko's research team from the Department of Artificial Intelligence at the University of Seoul has been accepted at IJCAI 2026, one of the most prestigious international conferences in the field of artificial intelligence.

(Left to right) Sangki Ko, Professor at University of Seoul, Seongmin Kim, Master's student at University of Seoul. University of Seoul

(Left to right) Sangki Ko, Professor at University of Seoul, Seongmin Kim, Master's student at University of Seoul. University of Seoul

View original image

This study addresses the "Programming by Example (PBE)" problem, which involves automatically generating regular expressions solely from examples. The core of the research is the introduction of a novel recursive AI framework called "RESYN," capable of effectively synthesizing complex regular expressions used in real-world settings.


Previous research on regular expression synthesis has relied on benchmark data with relatively simple structures, which has limited their ability to capture the complex patterns found in real environments.


To address this, the research team quantitatively analyzed the structural complexity of regular expressions in real-world data and proposed a new approach to solve these challenges.


In particular, the team developed a lightweight neural network model called "SET2REGEX," which features a divide-and-conquer-based synthesis structure that recursively decomposes the problem and incorporates permutation invariance of the input example sets.


Through this approach, they achieved high performance with far fewer parameters than existing large-scale models, and demonstrated superior results over previous methods and large general-purpose language models on complex real-world datasets.


Co-authors of the paper include Professor Sangki Ko, Master's student Sungmin Kim from the University of Seoul, and researcher Yoseop Han from Yonsei University.



The research team expects that their results can be expanded into a general methodology applicable not only to regular expression synthesis but also to various program synthesis tasks and the resolution of complex, structure-based artificial intelligence problems.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing