How Tawseef AI
Classification Works
Tawseef AI classification combines advanced automation with expert validation to deliver enterprise-grade accuracy at scale. Achieved accuracy is highly dependent on the quality of your provided product data.
Dual engine technology with ML models, guided standards, web verification, and iterative improvement.
What Makes Tawseef AI Different
Four key technologies working together for superior accuracy
Dual Engine: ML + Guided Standards
Our proprietary models classify items with high accuracy, which is greatly influenced by the quality of the input data. You can further refine by selecting the relevant UNSPSC or GS1 GPC segment/class, boosting precision and adding clear context to the final result.
- Proprietary ML models trained on millions of products
- Manual segment/class selection for added context
- Combines AI speed with human expertise
Web Verification for Part Numbers
We automatically fetch and verify details from the web (e.g., manufacturer part numbers), even when they're missing from the item description — raising confidence and fill-rate.
- Automatic web lookup for part numbers
- Enhanced confidence and data completeness
- Works even with minimal input data
Client-Specific Dictionaries
Optional, project-level custom dictionaries capture client terminology, supplier aliases, and local jargon for even better accuracy.
- Industry-specific terminology support
- Supplier alias mapping
- Localized jargon recognition
Actionable Reliability Score
Every classification includes a Reliability Score so teams can focus on low-confidence items first, maximizing efficiency and quality.
- Confidence scoring on every item
- Priority-based review workflow
- Focus resources where needed most
How It Works: Step by Step
Iterative process for continuous accuracy improvement
Initial Run: Ingest & Classify
Upload your product data (CSV, Excel, or via API). Our system ingests the data and generates initial classifications using our proprietary ML models.
Targeted Improvements: Review & Refine
Use the Reliability Score to identify weaker items. Add context like attributes, UNSPSC/GPC segment hints, and custom dictionaries to improve accuracy.
Re-Run: Lift Overall Accuracy
Reprocess the data with your improvements to lift overall accuracy, especially down to commodity level where rich descriptions matter most.
Full-Service: AI + Expert Supervision
For Full-Service engagements, our AI processes your data at scale (20K–100K items/hour). Tawseef experts then supervise the AI's output and validate low-confidence items, ensuring the highest quality results without manual bottlenecks.
Tips for Best Accuracy
- Provide rich descriptions: Especially important for commodity-level precision
- Use guided standards: Select UNSPSC/GPC segments to encode business context
- Leverage custom dictionaries: Add industry-specific terminology and aliases
- Iterate: Run → review low scores → add context → re-run for continuous improvement
Choose the model that fits your team and volume
API
For high-volume or real-time needs. Plug directly into ERP, PIM, or MDM and classify on demand.
- Real-time classification
- Direct system integration
- 20K–100K items/hour
Self-Service
Upload files through a secure portal and receive results back quickly. Ideal when you have internal capacity to iterate.
- File upload/download
- Client-managed iteration
- Dashboard & reporting
Full-Service
AI processes your data at enterprise scale. Tawseef experts supervise the AI and validate quality. Recommended for very large datasets or establishing a clean foundation.
- AI-driven at scale
- Expert supervision
- Strategic validation
Ready to Get Started?
Experience the power of intelligent classification