Value Clusterer & Standardizer

How to standardize text values

1. Drop a CSV, TSV, JSON, or spreadsheet file here, or click browse files. You can also paste a table into the text area.
2. Select columns with Select all, Select none, or the individual column checkboxes.
3. Choose a Similarity method, adjust Similarity threshold, and choose Replace in column or Write to new column.
4. Check each cluster, edit Canonical value when needed, then click Approve mapping or Reject (leave values as-is).
5. Click Export cleaned dataset (CSV) for the data, or Export mapping table (JSON) for approved mappings.

What the similarity methods change

Normalized text (trim, case, whitespace) trims values, collapses repeated whitespace, and lowercases them. Fingerprint (key collision, ignores word order) removes punctuation, removes repeated words, sorts the remaining words, and groups matching keys. Edit-distance (Levenshtein, catches typos) compares normalized values with the threshold. Phonetic (Soundex, catches sound-alikes) groups matching Soundex codes. For the first, second, and fourth methods, a threshold below 100% also compares representative values with normalized-text similarity.

city
New York
new york
 New York

The normalized method puts these three strings in one cluster. Set Canonical value to New York and click Approve mapping. The cleaned output can then contain:

city
New York
New York
New York

How input is parsed and output is written

CSV and TSV input uses the first row as headers. JSON accepts an array of records or an object containing an array of record objects. JSONL and NDJSON are read one JSON object per line. An XLSX or XLS file uses its first worksheet. Pasted text is read as JSON, JSONL, comma-separated data, or tab-separated data based on its contents.

Blank, null, and undefined values are skipped while distinct values are collected. A file with no rows cannot load, and files over 20,000 rows are rejected. Approved mappings change values in selected columns, or create a <column> (standardized) column. Rows remain in the output, while unreviewed and rejected values stay as-is.

Load the new dataset, then use Import mapping to reapply an exported JSON mapping. Matching entries appear for review again. The page parses, clusters, and exports in the browser, and its code makes no request carrying the dataset content.

Using the review controls

Select Write to new column when the original values need to remain available. Approved mappings are written to a <column> (standardized) column.

Approve mappings, export Export mapping table (JSON), and load another dataset. Use Import mapping to reapply the JSON mapping, then review matching entries again.

Use Edit-distance for spelling variations. Inspect source examples and occurrence counts, then click Reject (leave values as-is) when the values should stay separate.

When a different dataset tool fits

Use Value Clusterer & Standardizer when alternate values in selected columns need per-cluster review. Start with Dataset Clean Tools when the task includes several kinds of cleanup. Choose Normalize Null Values for missing-value markers with per-column targets. Use Remove Duplicate Rows from CSV, Excel, JSON & Tables when whole records repeat. For address fields, Address Format Standardizer handles field roles, layout choices, and incomplete-record flags.

Frequently asked questions