Explore your own data
Upload a file and every passage is scored with
(nine Likert dimensions), the narrative classifier, and WebOrganizer topic and
format. You then browse the results in the same explorer as the corpus.
Format
- .csv or .jsonl with a
textcolumn, one document per row. An optionalidcolumn is kept if its values are unique. Any other columns are shown on the cards and included in the CSV export. - .txt: one document per line.
- UTF-8, up to 5 MB.
What happens
- Documents are split into passages the same way as the corpus: sentence windows of 50 to 200 words. Shorter documents are kept and flagged.
- The first 500 passages are scored. This takes a few minutes.
- Scores are model predictions. The models were trained on English web text, so they are less reliable on other languages and very different genres.
Privacy
- Don't upload anything sensitive or personal.
- You get a private link. Your file and its scores are deleted after 48 hours, or sooner if you delete them.
- We don't use your text for research. We keep only counts (number of uploads, sizes, durations).
- Limit: 5 uploads per hour.