Load documents
Paste text, or drop a text/JSONL file with one document per line. Define entity types with a key and a colour.
Annotation
Entity (NER) Span Annotator
Select any run of text with the mouse and press the key for its entity type. Spans export in the spaCy JSONL shape (text plus start, end, label offsets), in BIO tag format tokenised on whitespace for Hugging Face token classification, and as CoNLL.
Loading the tool…
How it works
Tag spans
Select text, press the type's key. Click a span to change or remove it. Overlaps are rejected so exports stay valid.
Export
spaCy JSONL, BIO/IOB2 token lines, or CoNLL columns. Offsets are character-based in the original text.
Formats, limits and the numbers that matter
Tokenisation
BIO export splits on whitespace and punctuation; a span that starts or ends inside a token is snapped to token boundaries and flagged.
spaCy training
Convert the JSONL with spacy convert or a small script into .spacy DocBin; the offsets here are what doc.char_span expects.
Hindi and Urdu
Right-to-left text is supported for selection and export; offsets are in Unicode code points.
Common questions
Nested entities?
Not in this version; each character belongs to at most one span.
Can I pre-annotate with a pattern?
Yes: the regex box tags every match with a chosen type; review before exporting.
Can I import spaCy JSONL?
Drop the file and the spans are restored.