Messy Data Standardizer — Cluster Spelling Variants
Review spelling clusters, choose standards and rewrite only approved values.
Try before you buy.
- Free preview: the row and cluster counts and the first clusters and cleaned rows (up to 10 and at most half; a single result has its values masked).
- Locked until you unlock it: download, copy and saving to your device.
- Unlock: Pro pass, ₹179 for 30 days or ₹79 for 1 week, a one-time payment that never renews. Unlock: Pro pass, $5.99 for 30 days or $2.99 for 1 week, a one-time payment that never renews.
Ways to unlock shows how to get the full result.
Printing this result is locked in the free preview.
Table and target column
A selected file is used instead of pasted text. Workbook formulas are not recalculated; displayed or cached cell values are read as text.
Clustering rules
Methods: fingerprint, ngram, soundex, double-metaphone or nearest. Edit legal suffix words and literal substitutions, case, accents and punctuation. ngram chooses the piece size; threshold chooses the minimum nearest-neighbour score.
Clusters are suggestions. Fingerprints ignore word order; phonetic methods approximate Latin-script names. Nearest matching uses blocking for large columns and can miss pairs. Review every member before merging.
Choose only merges you approve
Cleaned table and reusable mapping
The free preview shows counts and part of the clusters and cleaned table. A pass opens every cluster and the whole cleaned file, mapping and exports. Saved rules are settings and are never locked.
Locked in the free preview. Opens the ways to unlock this result.
Locked in the free preview. Batch runs unlock with a pass.
Locked in the free preview. Query results unlock with a pass.
About the Messy Data Standardizer — Cluster Spelling Variants
The same organisation or product may appear as several spellings in one column. This workspace groups likely variants, shows their original values and counts, and lets you choose or type one standard. Fingerprint-style word keys, n-gram keys, Soundex, Double Metaphone and blocked Levenshtein comparisons provide different review suggestions. Nothing is rewritten merely because two values share a key.
The free preview shows the counts and part of the clusters and cleaned table. A Pro pass opens every cluster and the whole cleaned file, reusable original-to-standard mapping, copying and exports. Literal abbreviations, suffix words and normalisation rules can be remembered locally without unlocking a result.
How to use it
- Open CSV, TSV or a supported workbook, or paste delimited cells. Check the delimiter, header setting and workbook sheet.
- Read the table and choose the column whose values need a standard spelling.
- Edit the rules object: fingerprint, ngram, soundex, double-metaphone or nearest, plus suffixes, literal substitutions and case or accent handling.
- Find clusters, inspect their members and counts, then type a standard and explicitly select each merge you approve.
- Apply the selected merges. With a Pro pass, copy or export the cleaned table and mapping; keep the original column if you want an audit trail.
Examples
Tata Motors Ltd TATA MOTORS LIMITED Tata Motors
One suggested cluster after the configured case and suffix rules. No cells change until you select a standard and apply the merge.
Common uses
- Standardising organisation names before joining exports.
- Reviewing spelling variants in survey or product columns.
- Keeping an original-to-standard mapping for a later import workflow.
How clustering suggestions work
A fingerprint sorts and deduplicates normalised words. An n-gram fingerprint sorts distinct character pieces. Phonetic methods approximate Latin-script names and retain numeric or uncoded tokens separately. Nearest comparisons use a Levenshtein score and candidate blocking for large lists. The OpenRefine clustering reference explains these families and their review tradeoffs.
Explicit merges and source preservation
Every cluster initially remains unselected. The most frequent original spelling is suggested, with the first source occurrence breaking ties, but you can type any standard. Applying a selected mapping changes only exact original values. Keeping the original column adds a separate standard column; otherwise the input is still retained in this page’s local run.
Phonetic and nearest methods
Soundex follows the name-key family described by the US National Archives. Double Metaphone supplies primary and secondary Latin-name codes. These keys compare spelling or approximate sound, not meaning. Nearest clusters are transitive: two distant members can be linked by an intermediate value. Blocking can miss candidate pairs, so every proposed merge needs review.
Limitations
- Clustering does not prove identity. Different organisations, people or products can share a key.
- Device memory, long cells and candidate comparisons affect practical size. Nearest-neighbour matching takes more work than fingerprint keys; split unusually large or long-value columns.
- Values above 20,000 characters or columns above 200,000 distinct values require smaller local parts. Table reads use a bounded device working buffer.
- Workbook formulas are not recalculated. Cached or displayed values are read as text.
- Exports protect formula-prefix strings with an apostrophe and preserve identifiers as text in Excel. CSV consumers may infer types on import.
Privacy
On this device: nothing is sent
The table is read and clustered locally in a background thread; it is not uploaded. Only clustering rule settings are saved when you choose Remember these rules. Original input and generated mappings remain in this page’s run and are cleared on Reset or navigation.
Frequently asked questions
What do I get without a pass?
Without a pass, Messy Data Standardizer — Cluster Spelling Variants shows the row and cluster counts and the first clusters and cleaned rows (up to 10 and at most half; a single result has its values masked). Until you unlock it, the result can’t be downloaded, copied or saved to your device. A Pro, Premium or Ultimate pass, a one-time payment that never renews, unlocks the full result. The pricing page lists the passes and their prices.
Will a suggested cluster change my file automatically?
No. Review the variants, choose a standard and select Apply this reviewed merge. Only selected mappings are applied.
Which method should I start with?
Start with fingerprint for case, punctuation and word-order differences. Use n-grams for character-piece variants, phonetic methods for Latin names, or nearest matching for explicit spelling-distance comparisons.
Can I keep the original spelling?
Yes. Keep the original column is selected by default and adds a separate standard column. The mapping export records original and standard values.
Can nearest matching miss a pair?
Yes. Candidate blocking limits comparisons for large columns. Its warning remains visible, and a cluster’s transitive grouping is a review suggestion rather than an identity claim.