Sim WooChul
/

Sim WooChul.

I research and build AI systems that understand language.

Sim Woochul
About

Sim Woochul : AI Research Engineer with over seven years of experience researching and building language systems. I specializes in Korean NLP, patent retrieval, and transformer-based models, focusing on bridging research with real-world search and language systems.

Experience
2019 – Present

Research & Development

NLP / Patent Search

2015 – 2019

Development

Research Interests
01

Patent Retrieval & Semantic Search

02

Korean NLP & Tokenization

03

Transformer-based Models & LLMs

04

Name Disambiguation / Entity Resolution

Selected Work
2019Research on an Automated IPC Classification Recommendation System for Korean Patents
2020Research on Korean Patent Language Models and NER for Composition and Physical Properties in Chemical Patents
2021Research and Development of an Automated CPC Classification Recommendation System for Korean Patents
2022Research on a Weak Signal Detection System for Automatically Identifying High-Value Patents
2023Research and Development of an AI-based Trademark Name Similarity Search System
2024Research on Performance Enhancement of an Automated CPC Classification Recommendation System
2025Research on Constructing Korean Search Training Datasets to Improve AI-based Patent Search Systems
Tech Stack
LanguagePython
ML / DLPyTorch / HuggingFace / JAX
NLPCustom Korean Tokenizer / BERT-based Models
SearchFAISS / Dense Retrieval
InfraLinux / Docker
Research

Research

2022Best Paper Award

Korean Patent CPC Auto-Classification Using KorPatBERT

Pre-trained KorPatBERT on 120GB patent corpus; achieved 78.63% subclass accuracy and 86.57% Five Guesses maingroup accuracy on 2M+ documents.

Sim WooChul, Park JinWoo, Lee SangHeon, Ko BongSu, Noh HanSung
Journal of Intellectual Property, Vol.17 No.3

KorPatBERTCPCPatentNLP
2023

Future Technology Weak Signal Analysis Using Patent Keywords

Extracted 61.5M keywords and keyphrases from 500K CPC-A patents using KorPatBERT embeddings, detecting 74 weak signals via graph-based clustering.

Sim WooChul
KIPI Internal Research

Weak SignalPatentNLPKeyphrase
2023

Weak Signal Analysis and Growth Prediction Using Patent CPC Codes

Extracted 3,838 weak signals from 3.2M patents using CPC subgroup vectors; GCNN model achieved 97.44% accuracy predicting 10-year high-growth technologies.

Sim WooChul
KIPI Internal Research

Weak SignalCPCPatentGCNN
Contact

GET IN
TOUCH.

Feel free to reach out if you're interested in joint research or projects in NLP, Patent Search, or Deep Learning.

Name
Affiliation (optional)
Email
Inquiry Type
Message