Upload Survey Data
Drop your Excel or CSV file here
or browse files · .xlsx / .xls / .csv
Codebook Upload Optional Map OE responses to predefined codes
Drop Codebook here or browse files
Required columns: Code · Label · Keywords (comma-separated) · .xlsx/.csv
Off
No codebook loaded — auto-coding will use built-in clustering only.
Configure Dictionaries — English
90%

Jaccard word overlap above this = Similarity flag

Uses franc-min — fast offline trigram model (~50KB).

30%

If % of words found in the language dictionary is below this → Garbage. Each language has a sensible default (EN: 30%, others: 15%). Texts ≥ 25 chars are exempt. Set to 0% to use the language default only.

15%
🟣 AI Bot Detection Settings NEW

Detects AI-generated / LLM-written survey responses using 10+ linguistic signals from the Screen Engine AI Pattern Dictionary. Runs automatically during Scan — no extra step needed.

Active

Low = threshold 65 | Medium = 45 | Aggressive = 28

Short responses below this word count are skipped (AI rarely writes short).

Detection Signals (click to enable/disable)
Test a response
→ AI Score: —
Scanning…
🧠 AI Review Agent Running…
0 reviewed 0 flagged
No data yet
Upload a file or load a sample dataset, then click Scan Now.
Coding Engine Settings
70%

Jaccard similarity ≥ threshold → same cluster. Lower = bigger, broader clusters.

Forces smaller threshold to stay under limit. Useful for large surveys.

Semantic Normalization Rules NEW

Rules run before fuzzy matching. Matching responses are collapsed to a canonical key, guaranteeing they land in the same cluster regardless of phrasing.

Built-in Semantic Rules
Custom Rules (one per line: CANON_KEY = keyword1, keyword2, …)

Format: MY_KEY = phrase one, phrase two — if the cleaned response contains ANY listed phrase, it maps to MY_KEY.

Test a response
→ Canonical: —
Coding…
No data yet
Upload a file or load the Coding sample, then click Run Auto-Coding.