A research survey · 2026
Knowledge-Guided
Data Cleaning
From knowledge acquisition
to reliable decisions
What evidence supports a change to your data? Thirty studies connect the knowledge we acquire with the decisions we can trust.
Change the city or change the ZIP?
A conflict alone does not tell us which value is wrong.
Illustrative example · select the evidence to explore the decision
The central question
Detecting a conflict is only
the beginning.
A rule can expose an inconsistency. Choosing a correction requires something more: evidence that applies to the right record, attribute and task.
This survey follows that connection from knowledge acquisition to repair selection, then into multi-step data preparation. It distinguishes factual correctness, constraint consistency and downstream task utility.
Acquire knowledge
Where do rules, facts, patterns and feedback come from?
02Choose a repair
How does evidence become a specific decision?
03Establish reliability
When should a system change, revise or abstain?
04Prepare for a task
How do meanings and operation order shape the result?
“What were our actual sales
by month and city last year?”A simple request, a richer problemCorrect values are part of the answer. The workflow also needs to interpret “actual sales,” subtract refunds and coordinate transformations with joins.
The literature map
Thirty studies.
One connected conversation.
Explore the methods by the decision they support. Expand a study for a concise explanation and its presentation page.
Loading the literature map…
30 studies in the survey
Take home
Keep what works.
Recheck
what changes.
Every task should make the next one smarter. Reuse knowledge with the evidence that makes it trustworthy.
- 01
Remember what you verify
Retain verified rules, meanings and experience for later decisions.
- 02
Update with evidence
Revise or retire knowledge when its conditions no longer hold.
- 03
Ask when evidence is missing
Inspect the data, retrieve context or ask the user.
- 04
Carry knowledge across tasks
Connect cleaning, transformation and analysis.
- 05
Verify every step
Check how each operation contributes to the intended result.
Read, present, explore
Take the survey
with you.
The complete paper and presentation,
in formats ready to read and share.
The survey paper
From Knowledge Acquisition to Reliable Decisions
The presentation
Examples, method diagrams and take-home messages
Zhou, Y. (2026). Knowledge-Guided Data Cleaning: From Knowledge Acquisition to Reliable Decisions. Hong Kong Baptist University.