← BACK TO PROJECTS
Onpoint Insights

Matching Messy Company Names Across Datasets

Partial company names across order datasets created reporting inconsistencies for a packaging client.

→ Significant reduction in manual validation time
HOW THE SYSTEM WORKS
IN
Messy company names
partial and inconsistent across datasets
STEP
Entity recognition
finds the company in the text
STEP
Similarity matching
cosine similarity to canonical names
OUT
Clean reporting
consistent entities in Power BI
“IBM”, “I.B.M. Corp” and “IBM Corporation” become one clean entity.
Built
NLP-driven fuzzy matching system to reconcile partial entities (think 'IBM' vs 'I.B.M. Corp').
Approach
Named Entity Recognition + cosine similarity, integrated with Power BI dashboards.
Impact
Significant reduction in manual validation time
Skills
NERCosine SimilarityPower BINLP