GUM

Georgetown University Multilayer Corpus for NER | Entity Extraction

Description

The GUM corpus was collected and annotated at Georgetown University.

For more information, see the LICENSE, and the following publication: Zeldes, Amir (2016) "The GUM Corpus: Creating Multilayer Resources in the Classroom". Language Resources and Evaluation.

This corpus version has been modified from the original one to focus on NER about a set of 22 classes, and transformed into IOB2/CoNLL like format.

Entries

train: 44111 entries | test: 18236 entries

URL

https://github.com/amir-zeldes/gum

File Format

text - CoNLL - IOB2

Column	Description
token	a string feature
ner_tag	a classification label, 23 classes

Example

    Pacific	B-person
    Standard	I-person
    owner	I-person
    ,	O
    Jonathan	B-person
    M.	I-person
    Stan	I-person
    ,	O
    displays	O
    the	O
    Santorum	B-substance
    cocktail	I-substance
    drink	I-substance
    as	O
    a	B-object
    finished	I-object
    product	I-object