Skip to content
Original tutorialPublic source reviewed

Check CSV Blanks and Shifted Columns Before Jev Classification

Produce a preflight report with stable row IDs and separate input errors from model issues.

Use caseCheck missing values, duplicates, and column shifts before classifying a support CSV

Source · JevLog originalBeginner6 min

Inspect a CSV with ID and text columns before classification. Make a local report, fix blank or shifted fields, and write a separate copy only if needed. Keep the original file and verify row IDs.

Source · JevLog original · Original guide

2026-09-21 · Editorial update; no live model call

Blank text, repeated exports, and shifted columns can change a classification. Inspect a small working copy with the local Python script, then check its row and column report. Fix structural errors before writing a new CSV. The script does not use the network or delete rows.

  • Python 3.10 or newer and a UTF-8 CSV, TSV or semicolon-delimited file.
  • Know the name of the column containing your text.

Check CSV columns, blanks, duplicates, and row shape before classification. AI concept art.

Check CSV columns, blanks, duplicates, and row shape before classification. AI concept art.

Export only stable IDs and needed text; keep the original file. Check the header and a few rows before inspecting.

2. Run inspection without creating an output

Section titled “2. Run inspection without creating an output”

Run inspection on the bundled example, then your small file. Check row counts, columns, blanks, and repeats; no new CSV is written yet.

Terminal window
python scripts/prepare_csv.py examples/customer_feedback.csv --text-column feedback

Fix duplicate or missing columns and uneven rows at the source, then inspect again. Python’s CSV parser handles quoted commas and newlines.

Check IDs behind repeated text; identical words may be separate customer records. The report flags repeats without deleting them.

Use –output for a new copy, then verify rows, IDs, and text in Studio. Check spreadsheet exports for formula-like cells.

Terminal window
python scripts/prepare_csv.py examples/customer_feedback.csv --text-column feedback --output clean_feedback.csv

Practice sample: download customer_feedback.csv (CSV Studio / local practice; not a live run here).

You get a JSON report and optional CSV copy traceable by ID. Parsing succeeds; classification still needs review.

  • You can identify the input, its source, and the fields that leave your system.
  • Original identities, failed rows, uncertain cases and human corrections remain visible.
  • You distinguish an offline fixture, an author demo and a live evaluation you ran yourself.

No. Export real UTF-8 CSV or TSV from the source application.