Harmonized Product Classification via Hybrid AI Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current manual classification of products using the Harmonized System (HS) is inefficient and prone to errors, leading to incorrect tax applications and potential legal and financial consequences.
Innovation Solution
The implementation of a hybrid deep learning and generative AI system that uses multimodal classifiers, including deep learning models, large language models, and semantic similarity models, to automate the classification of products and provide accurate HS code suggestions with confidence scores and rationales.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual classification is used, then flexibility in interpretation is maintained, but accuracy and consistency deteriorate
Solution Approach 1:
The classification system is segmented into multiple specialized models: a deep learning model for extracting product features, a generative AI model for handling unstructured data, and a voting model for aggregation. Each model focuses on specific aspects of classification, improving overall accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
Multiple AI models with different strengths are merged into a unified classification system. The deep learning model processes structured data, the generative AI model handles unstructured descriptions, and their predictions are combined through voting. This merging leverages complementary capabilities to achieve higher reliability than any single model could provide.
2Measurement precision
If multiple AI models are used, then classification accuracy is improved, but computational complexity increases
Solution Approach 1:
A voting model acts as an intermediary between the deep learning model and generative AI model. It aggregates their predictions through a standardized interface, coordinating their outputs without requiring direct integration. This intermediary simplifies the overall architecture by providing a clear separation of concerns and standardized communication protocols.
3Reliability
If deep learning and generative AI models are combined, then classification robustness is improved, but processing time increases
Solution Approach 1:
The system implements a tiered processing approach where the voting model can operate with partial inputs from available models. If one model completes processing slower, the voting mechanism can still produce results based on available predictions, reducing overall processing time while maintaining robustness through the ensemble approach.
4Productivity
If automated classification is implemented, then efficiency is improved, but interpretability deteriorates
Solution Approach 1:
The system incorporates feedback mechanisms where the voting model aggregates predictions from multiple AI models and provides confidence scores. This feedback loop allows the system to explain its decisions by showing which models contributed to the final classification and with what confidence, maintaining interpretability while achieving automated efficiency.
Data Source
AI summary
This disclosure provides systems, methods, and devices for automatic product classification using deep learning and generative artificial intelligence (AI) models for a harmonized system (HS) product classification. A method includes generating embeddings based on an input dataset. The method includes applying a deep learning model to the embeddings to produce a prediction set including classifications corresponding to the embeddings. The method includes converting the classifications to a first set of similarity metrics. The method includes determining a second set of similarity metrics based on the embeddings using a semantic similarity model. The method includes generating a third set of similarity metrics based on an output of the semantic similarity model. The method includes outputting a ranked set of predictions corresponding to the input dataset based on the first set of similarity metrics, the second set of similarity metrics, and the third set of similarity metrics.