metadata

annotations_creators:
  - expert-generated
language:
  - hu
language_creators:
  - other
license: []
multilinguality:
  - monolingual
paperswithcode_id: null
pretty_name: SzegedNER
size_categories:
  - 1K<n<10K
source_datasets:
  - original
tags:
  - hungarian
  - szeged
  - ner
task_categories:
  - token-classification
task_ids:
  - named-entity-recognition

Introduction

The recognition and classification of proper nouns and names in plain text is of key importance in Natural Language Processing (NLP) as it has a beneficial effect on the performance of various types of applications, including Information Extraction, Machine Translation, Syntactic Parsing/Chunking, etc.

Corpus of Business Newswire Texts (business)

The Named Entity Corpus for Hungarian is a subcorpus of the Szeged Treebank, which contains full syntactic annotations done manually by linguist experts. A significant part of these texts has been annotated with Named Entity class labels in line with the annotation standards used on the CoNLL-2003 shared task.

Statistical data on Named Entities occurring in the corpus:

       | tokens | phrases
------ | ------ | -------
non NE | 200067 | 	 
PER    | 1921   | 982
ORG    | 20433  | 10533
LOC    | 1501   | 1294
MISC   | 2041   | 1662

Reference

György Szarvas, Richárd Farkas, László Felföldi, András Kocsor, János Csirik: Highly accurate Named Entity corpus for Hungarian. International Conference on Language Resources and Evaluation 2006, Genova (Italy)

Criminal NE corpus (criminal)

The Hungarian National Corpus and its Heti Világgazdaság (HVG) subcorpus provided the basis for corpus text selection: articles related to the topic of financially liable offences were selected and annotated for the categories person, organization, location and miscellaneous. There are two annotated versions of the corpus. When preparing the tag-for-meaning annotation, our linguists took into consideration the context in which the Named Entity under investigation occurred, thus, it was not the primary sense of the Named Entity that determined the tag (e.g. Manchester=LOC) but its contextual reference (e.g. Manchester won the Premier League=ORG). As for tag-for-tag annotation, these cases were not differentiated: tags were always given on the basis of the primary sense.

Statistical data on Named Entities occurring in the corpus:

       | tag-for-meaning | tag-for-tag
------ | --------------- | -----------
non NE | 200067          | 	 
PER    | 8101            | 8121
ORG    | 8782            | 9480
LOC    | 5049            | 5391
MISC   | 1917            | 854

Metadata

dataset_info:

config_name: business features:
- name: id dtype: string
- name: tokens sequence: string
- name: ner_tags sequence: class_label: names: 0: O 1: B-PER 2: I-PER 3: B-ORG 4: I-ORG 5: B-LOC 6: I-LOC 7: B-MISC 8: I-MISC
- name: document_id dtype: string
- name: sentence_id dtype: string splits:
- name: original num_bytes: 4452207 num_examples: 9573
- name: test num_bytes: 856798 num_examples: 1915
- name: train num_bytes: 3171931 num_examples: 6701
- name: validation num_bytes: 423478 num_examples: 957 download_size: 0 dataset_size: 8904414
config_name: criminal features:
- name: id dtype: string
- name: tokens sequence: string
- name: ner_tags sequence: class_label: names: 0: O 1: B-PER 2: I-PER 3: B-ORG 4: I-ORG 5: B-LOC 6: I-LOC 7: B-MISC 8: I-MISC
- name: document_id dtype: string
- name: sentence_id dtype: string splits:
- name: original num_bytes: 2807970 num_examples: 5375
- name: test num_bytes: 520959 num_examples: 1089
- name: train num_bytes: 1989662 num_examples: 3760
- name: validation num_bytes: 297349 num_examples: 526 download_size: 0 dataset_size: 5615940