Related Experiment Video
Updated: Jan 9, 2026
09:11
Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence
Published on: January 27, 2023
2.7K
MBCS: A few-shot intent detection model for manual inspection records
Mengjie Liao1,2, Yixin Wang3, Jian Zhang1,2
1School of Management Science and Engineering, Beijing Information Science and Technology University, Beijing, China.
Plos One
|December 10, 2025
Summary
This study introduces MBCS, a novel model for cargo risk intent detection that overcomes annotation scarcity. It significantly improves accuracy in few-shot scenarios, offering a robust solution for import-export trade supervision.
Area of Science:
- Natural Language Processing
- Machine Learning
- Artificial Intelligence
Background:
- Annotation scarcity in manual inspection hinders cargo risk intent detection accuracy.
- Few-shot learning scenarios present challenges in domain-specific data and text representation.
Purpose of the Study:
- To propose an intent detection model, MBCS (Multi-task Learning with BERT for Classification and Semantic Similarity Comparison), for few-shot scenarios.
- To enhance semantic representation capabilities and address annotation scarcity in cargo risk intent detection.
Main Methods:
- Implemented a multi-task learning framework combining text classification and semantic similarity comparison.
- Incorporated semantic contrastive learning as an auxiliary task to boost semantic representation.
- Utilized an attention-weight-based synonym substitution strategy for improved contextual understanding.
Main Results:
- MBCS achieved accuracy improvements of over 4.19% (5-shot) and 4.91% (10-shot) on real-world customs datasets.
- Demonstrated substantial performance gains compared to baseline models in few-shot intent detection.
- Validated the effectiveness of the multi-task learning framework and synonym substitution strategy.
Conclusions:
- MBCS offers an optimized solution for intent detection tasks facing annotation scarcity.
- The proposed model significantly enhances accuracy and generalization in cargo risk assessment.
- This research contributes a practical approach to improving import-export trade supervision through advanced NLP techniques.
