Handwriting & document AI
Recognition and structural analysis of handwritten mathematical expressions, Japanese/English handwriting, chemical formulas, and document-like input.
Assistant Professor / Applied AI Researcher
Building AI systems that understand handwriting, documents, and visual structure.
I am an Assistant Professor at the WACOM-TUAT Joint Research Lab, Tokyo University of Agriculture and Technology. My work focuses on machine learning, computer vision, handwriting recognition, automatic scoring, image/text understanding, recommendation-oriented applications, and deployable data-driven AI systems.
Graph-based reasoning
Graph neural networks for modeling symbols, spatial relations, candidate edges, and structural patterns in handwritten mathematical expressions.
Applied AI systems
End-to-end model development, REST API deployment, computer vision systems, social media analytics, recommendation systems, and real-world AI product prototypes.
Current research directions
Graph-based HME structure recognition
I develop models that represent handwritten mathematical expressions as structured graphs of symbols and spatial relations, with edge-aware attention and relation prediction.
PyTorch GeometricGATCROHMEInkML
Multimodal handwriting recognition
I combine online pen trajectories, stroke order, and timestamps with offline visual representations to build more robust handwriting and mathematical expression recognition systems.
Online signalsOffline imagesMultimodal AI
Automatic scoring of handwritten answers
I work on recognition, answer similarity, confidence estimation, rejection mechanisms, and human-in-the-loop verification for educational assessment.
Educational AIScoringHuman-in-the-loop
Geometry drawing assessment
I study process-aware scoring of geometric construction tasks using digital ink from an electronic drawing compass, stroke-level clustering, step-level segmentation, primitive extraction, and geometric reasoning.
Digital inkGeometryStep-wise scoring
Handwriting input systems for LMS
I develop Moodle-based handwriting input and recognition systems for math, chemistry, Japanese, and English questions while preserving handwriting traces for analysis.
MoodleREST APIMathJaxEducational assessment
Teaching experience
Jan 2016 – Spring 2018
Teaching Assistant
Faculty of Computer Science & Engineering, Ho Chi Minh City University of Technology · Vietnam
- Spring 2018: Data Structures and Algorithms
- Fall 2017: Computer Graphics
- Spring 2017: Data Structures and Algorithms
- Fall 2016: Computer Graphics
- Spring 2016: Programming Fundamentals
Papers
Journal and conference papers in handwriting recognition, document AI, automatic scoring, and educational technology.
Improvement of End-to-End Offline Handwritten Mathematical Expression Recognition by Weakly Supervised Learning
A Survey on Handwritten Mathematical Expression Recognition: The Rise of Encoder-Decoder and GNN Models
ICDAR 2023 CROHME: Competition on Recognition of Handwritten Mathematical Expressions
Global Context for Improving Recognition of Online Handwritten Mathematical Expressions
Syntactic Data Generation for Handwritten Mathematical Expression Recognition
A Transformer-Based Math Language Model for Handwritten Math Expression Recognition
Relation-Based Representation for Handwritten Mathematical Expression Recognition
Online Handwritten Mathematical Symbol Segmentation and Recognition with Bidirectional Context
Learning Symbol Relation Tree for Online Handwritten Mathematical Expression Recognition
Medical Images Sequence Normalization and Augmentation: Improve Liver Tumor Segmentation from Small Dataset
Content-Based Similarity for Automatic Scoring of Handwritten Descriptive Answers
Two Experiments for Automatic Scoring of Handwritten Descriptive Answers
Automatic Scoring System for Digitized Handwritten Answers
Hierarchical Stroke-Level Clustering and Step-Level Segmentation for Automatic Scoring of Geometric Construction Answers with an Electronic Drawing Compass
Automated Recognition and Scoring of Handwritten Short Answer: Insights from Japanese Elementary and Junior High Schools
Enhancing LMS-Based Assessment with Handwriting Input: The Design and Implementation of HAIMR
A Moodle Plugin for Handwritten Answer Input, Management, and Recognition Supporting Japanese, English, Mathematical, and Chemical Formulas
Featured projects
Edge-aware GAT for HME
Relation prediction between handwritten mathematical symbols using edge-aware attention and relation decoding.
HAIMR Moodle Plugin
Handwritten answer input, management, and recognition for Japanese, English, mathematical expressions, and chemical formulas.
E-compass scoring
Process-aware scoring framework for geometry construction problems using arcs, lines, intersections, and step evidence.
Personal interests
Outside research, I enjoy staying active and documenting everyday moments.
Football
Badminton
Running
Photography
Cycling
Research vision
My long-term goal is to develop AI systems that can understand human-generated visual and multimodal information, including handwriting, documents, diagrams, and learning processes. I am particularly interested in combining computer vision, graph-based structural reasoning, online/offline handwriting recognition, and multimodal learning to build reliable and interpretable AI systems. My research aims to bridge academic advances in pattern recognition with practical applications in education, document intelligence, and human-centered AI.
I aim to bridge academic advances in pattern recognition with practical applications in education, document intelligence, and human-centered AI.