AI Briefing[CLEF 2022] Context-Aware Named Entity Recognition and Relation Extraction - LG AI Research Blog
Key point
LG AI Research demonstrated its technical capabilities by achieving high performance in the ChEMU 2022 challenge, which extracts information from chemical patents.
Details
LG AI Research is researching Information Extraction technology in the chemical field in collaboration with the world-renowned publisher Elsevier. Technology that converts chemical experiment and substance information from unstructured text into structured data plays a key role in reducing chemists' research time.
In the ChEMU 2022 challenge, LG AI Research participated in two Tasks (1a, 1b) that extract reaction information from chemical patents, achieving the highest performance among participating teams. The key achievements are as follows.
- Building a Domain-specific language model: Developed a language model specialized for the chemical field by Post-training Bio-LinkBert using chemical patent (23GB) and journal (22GB) data.
- NER and RE model optimization: Adopted a Sequence tagging approach to secure high performance regardless of Entity length, and maximized performance by configuring the model as a Pipeline rather than a single model.
Task 1a aims at Named Entity Recognition (NER), which extracts chemical substances and numerical information, while Task 1b aims at Event Extraction, which extracts relationships between trigger words and entities.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.