NER & Beyond
User Manual; Results

Annotation tagset mapping and transliteration

Description of a procedure for harmonisation of two annotation tag sets:
  1. User uploads a set of source annotated documents with any tagset with one or more document with gold (target) annotation tagset
  2. System reads both tagsets and offer mappings (for each tag, some option should be selected)
  3. System generates source documents in a new annotation scheme
  4. User downloads the results
Upload a .zip file with following structure:
f.zip
├── gold/
└──── rnd1.ann
└──── rndm2.ann
├── to_map/
└──── x1.ann
└──── x2.xml
└──── x3.txt
Download Sample Archive
Select an archive:

BRAT → CoNLL02 (.py source)

Upload a .zip file with following structure:
f.zip
├── f1.ann
├── f1.txt
├── f2.ann
└── f2.txt
Download Sample Archive
Select an archive:
Select tokenizer:

CONLL02 → BRAT (.py source)

Upload a .zip file with following structure:
── f.zip
├── f1.conll
└── f2.conll
Download Sample Archive
Select an archive:

BRAT → XML

Upload a .zip file with following structure: Please, remove attributes from ann file, if any
files.zip
├── f1.ann
└── f1.txt
└── f2.ann
└── f2.txt
Download Sample Archive
Select an archive:

XML → BRAT

Upload a .zip file with following structure:
files.zip
├── f1.xml
└── f2.xml
Download Sample Archive
Tags to be ignored (separated by spaces):
Tags that contain NEs (separated by spaces):
Select an archive:

XML → CONLL02

Upload a .zip file with following structure:
files.zip
├── f1.xml
└── f2.xml
Download Sample Archive
Select an archive:

Perfom NER with spaCy (output in BRAT and XML formats)

Upload a .zip file with following structure:
── f.zip
├── f1.txt
└── f2.xml
Download Sample Archive

Information about models: English, Serbian, Spanish, German,
Portuguese, French, Italian, Dutch, Multi Language.

Visualisation & Automatic Annotation
Select an archive:
Replace < and > with ᐸ and ᐳ for XML files
Select language model:

Perfom NER with StanfordNER (output in XML format)

Upload a .zip file with following structure:
── f.zip
├── f1.txt
└── f2.xml
Download Sample Archive

Information about models: Serbian, English, German.
Select an archive:
Replace < and > with ᐸ and ᐳ for XML files
Select language model:

NER stats on .ann files

Upload a .zip file with following structure:
f.zip
├── eng/
└──── f1.ann
└──── f2.ann
├── slv/
└──── f1.ann
└──── f2.ann
Download Sample Archive
Select an archive:

Gemini tool for NER evaluation, GitHub

Upload a .zip file with the following structure:
f.zip
├── gold/
└──── f1.ann
└──── f2.ann
└──── f3.ann
├── to_eval/
└──── f1.xml
└──── f2.xml
└──── f3.xml
gold/ directory can contain both .xml and .ann files
to_eval/ directory can contain .xml and .ann files (also .txt but well-formed)
it is important that the files to be evaluted and their corresponding gold standards have the same name
if at least one file is XML, than the evaluation will produce visualisation (generated .html files)
Download Sample Archive
Matching type score:
Strict Weighted
Alignment:
Greedy Max
Visualise by annotation type (at least one file has to be XML, separated with spaces):
Other options:
Generate a CSV Compare annotations by each type



Select an archive: