# Crowdsourced taxonomy verification: A feedback-driven framework for refining knowledge graph relationships via online search interactions

[Grab](https://yomu.fyi/company/grab) · Junpeng Niu · Jul 30, 2026

## Summary

Maintaining accurate Knowledge Graphs in dynamic domains like e-commerce and food delivery is challenging because automated language models frequently hallucinate relationships while manual curation cannot scale. To validate structural taxonomy links continuously, a closed-loop verification framework operationalizes search interfaces by injecting unverified candidate edges as hypotheses into live user traffic. The system uses an exploration-exploitation strategy to place candidate relationships in lower-risk interface slots, tracking contextually anchored micro-interactions such as clicks, dwell times, and purchases. An offline verification engine aggregates these weighted interactions into normalized confidence scores, automatically promoting verified links to permanent graph edges and pruning refuted relationships.

## Takeaways

- Knowledge graph connections are partitioned into verified edges that power standard search and candidate edges that act as probabilistic hypotheses awaiting validation.
- Interaction scoring employs a weighted tier system that assigns greater confidence to high-intent actions like purchases and add-to-cart events compared to simple clicks or skips.
- Candidate edges are evaluated via a binary decision process that either cements them into the graph upon meeting a verification threshold or deletes them after falling below a rejection threshold.

**Tags:** [Architecture](https://yomu.fyi/topic/architecture), [LLMs](https://yomu.fyi/topic/llm), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Recommendation Systems](https://yomu.fyi/topic/recommendation-systems), [Search](https://yomu.fyi/topic/search)

[Read original post](https://engineering.grab.com/crowdsourced-taxonomy-verification)