YAML Metadata Warning:The task_categories "knowledge-graph" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

YAML Metadata Warning:The task_categories "entity-alignment" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

STEA Benchmarks: Spatio-Temporal Entity Alignment

STEA(W-I) and STEA(Y-I) are spatio-temporal entity alignment datasets, extending the STKGA [ICEWS↔Wikipedia] and [ICEWS↔YAGO] datasets with heterogeneous spatial information (GPS coordinates, textual location descriptions, administrative regions).

Dataset Description

Each dataset provides entity alignment between ICEWS coded event data and a knowledge graph (Wikipedia or YAGO), with spatial annotations at the triple level.

Formats

Spatial Triples (triples_1_spatial.tsv, triples_2_spatial.tsv):

head_id	tail_id	rel_id	time_id	flag	lat	lon	spatial_text	admin_region

lat/lon: WGS84 coordinates (empty if unavailable)
spatial_text: Textual location description
admin_region: Administrative region (country or sub-national)

Entity Locations (ent_locations_1.tsv, ent_locations_2.tsv):

entity_id	entity_name	lat	lon	spatial_text	admin_region

Spatial Coverage

Dataset	KG1 (ICEWS) GPS	KG2 GPS
STEA(W-I)	52.7%	14.4%
STEA(Y-I)	21.6%	4.7%

Data Sources

ICEWS: Harvard Dataverse ICEWS Coded Event Data (country-level geocoding via geonames)
Wikipedia: Wikidata SPARQL queries (P625 coordinates, P131 admin regions, P17 country)
YAGO: Cross-referenced to Wikipedia, then Wikidata for spatial attributes

Construction

The datasets were constructed by augmenting the STKGA temporal knowledge graph alignment datasets with spatial information:

Phase 1: Wikipedia entities queried against Wikidata for GPS coordinates (P625), administrative regions (P131), and country affiliations (P17)
Phase 2: YAGO entities bridged via Wikipedia URLs to Wikidata
Phase 3: ICEWS entities geocoded via country-name extraction matching against a curated gazetteer of ~200 countries and territories
Phase 4: Spatial data merged into triple files and entity location mappings
Phase 5: Quality validation (coordinate range checks, coverage statistics)

Statistics

STEA(W-I)

Split	Triples	Entities	Relations	Times
KG1 (ICEWS)	3,527,881	11,047	272	3,110
KG2 (Wikipedia)	198,257	15,831	226	-
Alignments	5,058 (1,518 train / 3,540 test)

STEA(Y-I)

Split	Triples	Entities	Relations	Times
KG1 (ICEWS)	4,192,555	26,863	272	1,934
KG2 (YAGO)	107,118	22,555	41	-
Alignments	18,824 (5,648 train / 13,176 test)

Citation

If you use these datasets, please cite both the original STKGA and STEA papers:

@inproceedings{stea2025,
  title     = {PhyWorld-Align: Heterogeneous Spatio-Temporal Entity Alignment},
  author    = {},
  booktitle = {},
  year      = {2025}
}

@inproceedings{stkga2024,
  title     = {STKGA: Spatio-Temporal Knowledge Graph Alignment},
  author    = {},
  booktitle = {},
  year      = {2024}
}

License

CC BY 4.0

Source

https://huggingface.co/datasets/eduzrh/STEA-benchmarks

Downloads last month: 40

Total file size:

296 MB