IBS Software created a bilingual Named Entity Recognition (NER) system utilizing Amazon Bedrock to process cargo logistics emails in English and Japanese. This solution achieves over 95% accuracy while reducing operational costs significantly, addressing challenges in manual interventions and accuracy trade-offs.
IBS Software's Cargo system handles thousands of bilingual email messages daily, essential for processing logistics data.
These messages require the extraction of various key information, including air waybill numbers, flight details, and weights, in both English and Japanese.
Building a robust Named Entity Recognition (NER) solution presented several challenges, primarily the need for accurate entity extraction in two languages while keeping costs manageable.
The complexity was increased due to manual interventions that slowed operational efficiency and a necessary trade-off between accuracy and cost.
IBS Software opted for Amazon Bedrock’s managed distillation capabilities to develop their NER solution.
By distilling knowledge from the Amazon Nova Pro into a more efficient Nova Lite model, they achieved an F1-Score accuracy of 95.085% and reduced operational costs by 14 times compared to previous implementations.
This project focused on extracting 23 different entity types vital in cargo logistics, addressing high accuracy and low latency requirements for real-time processing.
The successful deployment on AWS has allowed IBS Software to significantly improve operational efficiency in processing bilingual cargo logistics emails.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
IBS Software created a bilingual Named Entity Recognition (NER) system utilizing Amazon Bedrock to process cargo logistics emails in English and Japanese. This solution achieves over 95% accuracy while reducing operational costs significantly, addressing challenges in manual interventions and accuracy trade-offs.