Text Similarity Clustering

Upload a CSV/TXT or paste texts to find groups of similar items using SBERT embeddings.

Paste texts (one per line)

Or upload CSV/TXT

CSV: uses the specified column, or text if present, otherwise the first column. Up to 500,000 rows.

Clustering method

Finds texts that are nearly identical — reposts, copypasta, minor rewording. Good for deduplication. You set a similarity threshold; only texts above it are grouped together.

Text column

CSV column name to cluster on.

Similarity threshold

Higher = stricter matches.

Min cluster size

Output

Differentiator column (optional)

CSV column that distinguishes sources. Used to filter out clusters that are just one source repeating itself.

Min distinct differentiators

Clusters with fewer distinct values are hidden. Set to 2+ to require cross-source overlap.

Cluster within column (optional)

Clustering runs independently within each value. Pair with NEAR_SIM_EVENTS / URL_AMP_EVENTS from the temporal analyzer to find textually similar posts inside each flagged window.