Data · dataset · 2022
MultiTACRED - Multilingual TAC Relation Extraction Dataset
Listed in Hugging Face Datasets
Description
MultiTACRED is a multilingual version of the large-scale TAC Relation Extraction Dataset (nlp.stanford.edu/projects/tacred). It covers 12 typologically diverse languages from 9 language families, and was created by the Speech & Language Technology group of DFKI (dfki.de/slt) by machine-translating the instances of the original TACRED dataset and automatically projecting their entity annotations. For details of the original TACRED's data collection and annotation process, see the Stanford paper (aclanthology.org/D17-1004/).
Translations are syntactically validated by checking the correctness of the XML tag markup. Any translations with an invalid tag structure, e.g. missing or invalid head or tail tag pairs, are discarded (on average, 2.3% of the instances). Languages covered are: Arabic, Chinese, Finnish, French, German, Hindi, Hungarian, Japanese, Polish, Russian, Spanish, Turkish.
Read the rest (4 more)
Intended use is supervised relation classification. Audience - researchers. Please see our ACL paper (arxiv.org/abs/2305.04582) for full details.
NOTE: This Datasetreader supports a reduced version of the original TACRED JSON format with the following changes: - Removed fields: stanford_pos, stanford_ner, stanford_head, stanford_deprel, docid The motivation for this is that we want to support additional languages, for which these fields were not required or available. The reader expects the specification of a language-specific configuration specifying the variant (original, revisited or retacred) and the language (as a two-letter iso code).
The DatasetReader changes the offsets of the following fields, to conform with standard Python usage (see _generate_examples()): - subj_end to subj_end + 1 (make end offset exclusive) - obj_end to obj_end + 1 (make end offset exclusive) NOTE 2: The MultiTACRED dataset offers an additional 'split', namely the backtranslated test data (translated to a target language and then back to English). To access this split, use dataset['backtranslated_test'].
You can find the TACRED dataset reader for the English version of the dataset at huggingface.co/datasets/DFKI-SLT/tacred.
Links
Where it is published
- Hugging Face dataset page huggingface.co/datasets/DFKI-SLT/multitacred ↗
landing page · from Hugging Face
Documentation and papers
- arXiv:2305.04582 arxiv.org/abs/2305.04582 ↗
publication · from Hugging Face
Catalogue records · 1
- Hub API huggingface.co/api/datasets/DFKI-SLT/multitacred ↗
metadata API · from Hugging Face
Topics
- Stated by source
- text classification
- From keywords
- Computer Science & AI
- Inferred from text
- Audio 65%
Provenance · 1 source records, 10 field assertions
| Source | Key | Last seen | Raw |
|---|---|---|---|
| Hugging Face Datasets | DFKI-SLT/multitacred | 15 d ago | JSON v1 |
| Field | Assertion | Extractor | Evidence |
|---|---|---|---|
| access_level | source · Hugging Face | connector:huggingface@1.0.0 | /gated |
| concepts[field].local:field:computer-science-ai | mapping · Hugging Face | connector:huggingface@1.0.0 | |
| concepts[modality].local:modality:audio | enrichment · Hugging Face | keyword-concept-rules@1.0.0 | title+description (65%) |
| concepts[task].hf_task:text-classification | source · Hugging Face | connector:huggingface@1.0.0 | /tags[task_categories:*] |
| created_date | source · Hugging Face | connector:huggingface@1.0.0 | |
| description | source · Hugging Face | connector:huggingface@1.0.0 | /description |
| license_text | source · Hugging Face | connector:huggingface@1.0.0 | |
| publication_date | source · Hugging Face | connector:huggingface@1.0.0 | |
| title | source · Hugging Face | connector:huggingface@1.0.0 | /id |
| updated_date | source · Hugging Face | connector:huggingface@1.0.0 |