Datasets:

lmb-freiburg
/

entitynet

+{
+  "@context": {
+    "@language": "en",
+    "@vocab": "https://schema.org/",
+    "arrayShape": "cr:arrayShape",
+    "citeAs": "cr:citeAs",
+    "column": "cr:column",
+    "conformsTo": "dct:conformsTo",
+    "cr": "http://mlcommons.org/croissant/",
+    "data": {
+      "@id": "cr:data",
+      "@type": "@json"
+    },
+    "dataBiases": "cr:dataBiases",
+    "dataCollection": "cr:dataCollection",
+    "dataType": {
+      "@id": "cr:dataType",
+      "@type": "@vocab"
+    },
+    "dct": "http://purl.org/dc/terms/",
+    "rai": "http://mlcommons.org/croissant/RAI/",
+    "equivalentProperty": "cr:equivalentProperty",
+    "examples": {
+      "@id": "cr:examples",
+      "@type": "@json"
+    },
+    "samplingRate": "cr:samplingRate",
+    "extract": "cr:extract",
+    "field": "cr:field",
+    "fileProperty": "cr:fileProperty",
+    "fileObject": "cr:fileObject",
+    "fileSet": "cr:fileSet",
+    "format": "cr:format",
+    "includes": "cr:includes",
+    "isArray": "cr:isArray",
+    "isLiveDataset": "cr:isLiveDataset",
+    "jsonPath": "cr:jsonPath",
+    "key": "cr:key",
+    "md5": "cr:md5",
+    "parentField": "cr:parentField",
+    "path": "cr:path",
+    "personalSensitiveInformation": "cr:personalSensitiveInformation",
+    "recordSet": "cr:recordSet",
+    "references": "cr:references",
+    "regex": "cr:regex",
+    "repeated": "cr:repeated",
+    "replace": "cr:replace",
+    "sc": "https://schema.org/",
+    "separator": "cr:separator",
+    "source": "cr:source",
+    "subField": "cr:subField",
+    "transform": "cr:transform",
+    "containedIn": "cr:containedIn",
+    "excludes": "cr:excludes"
+  },
+  "@type": "sc:Dataset",
+  "conformsTo": "http://mlcommons.org/croissant/1.0",
+  "name": "entitynet",
+  "alternateName": [
+    "EntityNet",
+    "lmb-freiburg/entitynet"
+  ],
+  "description": "EntityNet is a large-scale image-text dataset harvested with the help of knowledge graphs for efficient training of CLIP-style vision-language models. Entities and their attributes are collected from the Wikidata knowledge graph (queried via the QLever SPARQL engine), turned into web image-search queries, and the resulting image-text pairs are downloaded from the web; text descriptions combine knowledge-graph attributes with the images' web alt-texts. The released version contains roughly 29.1M samples (around 33M images paired with about 46M texts) split into train, validation and test. To respect copyright the dataset does not redistribute image bytes: each sample stores the external image URLs together with the source web-page URLs, image dimensions and the query hashes used to retrieve it; users download the images themselves with the provided dataloader.",
+  "url": "https://huggingface.co/datasets/lmb-freiburg/entitynet",
+  "sameAs": [
+    "https://entity-net.github.io/",
+    "https://arxiv.org/abs/2505.02746"
+  ],
+  "license": "https://opensource.org/license/mit",
+  "inLanguage": "en",
+  "version": "0.32.1",
+  "datePublished": "2025-09-10",
+  "isLiveDataset": true,
+  "creator": [
+    {
+      "@type": "sc:Person",
+      "name": "Simon Ging"
+    },
+    {
+      "@type": "sc:Person",
+      "name": "Sebastian Walter"
+    },
+    {
+      "@type": "sc:Person",
+      "name": "Jelena Bratulić"
+    },
+    {
+      "@type": "sc:Person",
+      "name": "Johannes Dienert"
+    },
+    {
+      "@type": "sc:Person",
+      "name": "Hannah Bast"
+    },
+    {
+      "@type": "sc:Person",
+      "name": "Thomas Brox"
+    },
+    {
+      "@type": "sc:Organization",
+      "name": "University of Freiburg",
+      "url": "https://lmb.informatik.uni-freiburg.de/"
+    }
+  ],
+  "publisher": {
+    "@type": "sc:Organization",
+    "name": "University of Freiburg",
+    "url": "https://lmb.informatik.uni-freiburg.de/"
+  },
+  "citeAs": "@inproceedings{ging2025entitynet,\n  author    = {Simon Ging and Sebastian Walter and Jelena Bratuli{\\'c} and Johannes Dienert and Hannah Bast and Thomas Brox},\n  title     = {Using Knowledge Graphs to Harvest Datasets for Efficient {CLIP} Model Training},\n  booktitle = {Pattern Recognition, 47th {DAGM} German Conference, {DAGM} {GCPR} 2025, Freiburg, Germany, September 23--26, 2025, Proceedings},\n  series    = {Lecture Notes in Computer Science},\n  publisher = {Springer Nature Switzerland},\n  year      = {2026},\n  pages     = {287--302},\n  isbn      = {978-3-032-12840-9},\n  doi       = {10.1007/978-3-032-12840-9_19}\n}",
+  "keywords": [
+    "CLIP",
+    "vision-language",
+    "image-text",
+    "contrastive learning",
+    "knowledge graph",
+    "Wikidata",
+    "QLever",
+    "web-harvested",
+    "image-to-text",
+    "text-to-image",
+    "image-classification",
+    "English",
+    "10M - 100M",
+    "parquet"
+  ],
+  "dataCollection": "Entities are extracted from the Wikidata knowledge graph (accessed through the QLever SPARQL engine), filtered, and expanded with attributes; these are turned into web image-search queries. Image-text pairs are then downloaded from the open web. Only external URLs and metadata are stored, not image bytes.",
+  "dataBiases": "Content reflects what is indexed by web image search for Wikidata-derived queries and therefore inherits the biases, coverage gaps and noise of both Wikidata and the open web (e.g. over-representation of popular entities and Western/English-language web content).",
+  "personalSensitiveInformation": "Images and alt-texts are scraped from the public web and may incidentally contain personal data or depict identifiable people; bytes are not redistributed, only links. Wikidata may contain potentially harmful terms, which the authors filter during query construction.",
+  "distribution": [
+    {
+      "@type": "cr:FileObject",
+      "@id": "hf-repo",
+      "name": "hf-repo",
+      "description": "The EntityNet Hugging Face dataset Git repository.",
+      "contentUrl": "https://huggingface.co/datasets/lmb-freiburg/entitynet/tree/main",
+      "encodingFormat": "git+https"
+    },
+    {
+      "@type": "cr:FileObject",
+      "@id": "entitynet-entities",
+      "name": "entitynet-entities.parquet",
+      "description": "Knowledge-graph entities used to build the dataset: one row per Wikidata entity with its names/aliases, source provenance and taxonomy links.",
+      "containedIn": {
+        "@id": "hf-repo"
+      },
+      "contentUrl": "https://huggingface.co/datasets/lmb-freiburg/entitynet/resolve/main/entitynet-entities.parquet",
+      "encodingFormat": "application/x-parquet",
+      "sha256": "4ae13fbb15634111500d8b5388fc3179fd3849717e1cb2770ecbf91d217fe660"
+    },
+    {
+      "@type": "cr:FileObject",
+      "@id": "entitynet-queries",
+      "name": "entitynet-queries.parquet",
+      "description": "Web image-search queries derived from the entities; one row per query, indexed by query_hash, the value referenced by the `qh` column of the image samples.",
+      "containedIn": {
+        "@id": "hf-repo"
+      },
+      "contentUrl": "https://huggingface.co/datasets/lmb-freiburg/entitynet/resolve/main/entitynet-queries.parquet",
+      "encodingFormat": "application/x-parquet",
+      "sha256": "bc702c2722e220916f9811e0c206471d79d6f9b7e7f575cfe233ac8253f02126"
+    },
+    {
+      "@type": "cr:FileSet",
+      "@id": "images-metadata-files",
+      "name": "images-metadata-files",
+      "description": "Sharded parquet files holding the image-text sample metadata for the train, validation and test splits (the auxiliary minitrain debug shard is excluded).",
+      "containedIn": {
+        "@id": "hf-repo"
+      },
+      "encodingFormat": "application/x-parquet",
+      "includes": [
+        "images_metadata/train-*.parquet",
+        "images_metadata/val-*.parquet",
+        "images_metadata/test-*.parquet"
+      ],
+      "excludes": [
+        "images_metadata/minitrain-*.parquet"
+      ]
+    }
+  ],
+  "recordSet": [
+    {
+      "@type": "cr:RecordSet",
+      "@id": "splits",
+      "name": "splits",
+      "description": "The train/validation/test splits of EntityNet.",
+      "dataType": "cr:Split",
+      "key": {
+        "@id": "splits/name"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "splits/name",
+          "name": "name",
+          "description": "Split name (train, validation or test).",
+          "dataType": "sc:Text"
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "splits/url",
+          "name": "url",
+          "description": "Canonical URL identifying the split.",
+          "dataType": "sc:URL"
+        }
+      ],
+      "data": [
+        {
+          "splits/name": "train",
+          "splits/url": "cr:TrainingSplit"
+        },
+        {
+          "splits/name": "validation",
+          "splits/url": "cr:ValidationSplit"
+        },
+        {
+          "splits/name": "test",
+          "splits/url": "cr:TestSplit"
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "images",
+      "name": "images",
+      "description": "One record per dataset sample. A sample groups the image-text pairs retrieved for one search result and stores, as parallel lists, the external image URLs, source page URLs, image dimensions, candidate texts and the query hashes that produced them.",
+      "key": {
+        "@id": "images/key"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "images/split",
+          "name": "split",
+          "description": "Split this record belongs to (train/validation/test).",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "fileProperty": "fullpath"
+            },
+            "transform": [
+              {
+                "regex": "images_metadata/([a-z]+)-\\d+-of-\\d+\\.parquet$"
+              },
+              {
+                "replace": "val/validation"
+              }
+            ]
+          },
+          "references": {
+            "field": {
+              "@id": "splits/name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/key",
+          "name": "key",
+          "description": "Unique per-sample identifier string (e.g. 'test000000000'); also used as the WebDataset __key__ when building the image tars.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "key"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/qh",
+          "name": "qh",
+          "description": "List of query hashes (one per associated image) referencing the `query_hash` column of entitynet-queries.parquet; records which knowledge-graph-derived web search produced each image.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "qh"
+            }
+          },
+          "repeated": true,
+          "references": {
+            "field": {
+              "@id": "queries/query_hash"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/texts",
+          "name": "texts",
+          "description": "List of candidate text descriptions for the sample, combining web alt-text/titles with knowledge-graph attribute text.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "texts"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/images_link",
+          "name": "images_link",
+          "description": "List of external URLs pointing to the original image files on the web (image bytes are not redistributed).",
+          "dataType": "sc:URL",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "images_link"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/images_contextlink",
+          "name": "images_contextlink",
+          "description": "List of URLs of the source web pages on which each image was found.",
+          "dataType": "sc:URL",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "images_contextlink"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/images_width",
+          "name": "images_width",
+          "description": "List of image widths in pixels, aligned with images_link.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "images_width"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "images/images_height",
+          "name": "images_height",
+          "description": "List of image heights in pixels, aligned with images_link.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileSet": {
+              "@id": "images-metadata-files"
+            },
+            "extract": {
+              "column": "images_height"
+            }
+          },
+          "repeated": true
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "queries",
+      "name": "queries",
+      "description": "Web image-search queries generated from Wikidata entities (entitynet-queries.parquet). Linked from image samples via the `qh` column.",
+      "key": {
+        "@id": "queries/query_hash"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "queries/query_hash",
+          "name": "query_hash",
+          "description": "Stable hash identifying the query; referenced by the image samples' `qh` column.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "query_hash"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/entity",
+          "name": "entity",
+          "description": "Wikidata entity id (QID, e.g. 'Q144') the query was built from.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "entity"
+            }
+          },
+          "references": {
+            "field": {
+              "@id": "entities/entity"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/entity_name",
+          "name": "entity_name",
+          "description": "Human-readable name of the Wikidata entity.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "entity_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/query",
+          "name": "query",
+          "description": "The search query string sent to web image search.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "query"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/search_type",
+          "name": "search_type",
+          "description": "Image-search type filter used (e.g. 'Photo').",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "search_type"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/or_terms",
+          "name": "or_terms",
+          "description": "Additional OR-terms appended to disambiguate the query.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "or_terms"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/query_type",
+          "name": "query_type",
+          "description": "Query construction type (e.g. 'wd-noun-living').",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "query_type"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/attribute",
+          "name": "attribute",
+          "description": "Entity attribute the query encodes, if any.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "attribute"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/attribute_count",
+          "name": "attribute_count",
+          "description": "Number of attributes combined in the query.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "attribute_count"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "queries/attribute_category",
+          "name": "attribute_category",
+          "description": "Category of the attribute used.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-queries"
+            },
+            "extract": {
+              "column": "attribute_category"
+            }
+          }
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "entities",
+      "name": "entities",
+      "description": "Wikidata knowledge-graph entities the dataset is built from (entitynet-entities.parquet).",
+      "key": {
+        "@id": "entities/entity"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "entities/entity",
+          "name": "entity",
+          "description": "Wikidata entity id (QID, e.g. 'Q144').",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "entity"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/entity_name",
+          "name": "entity_name",
+          "description": "Human-readable entity name.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "entity_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/sitelinks",
+          "name": "sitelinks",
+          "description": "Number of Wikidata sitelinks, used as a popularity/quality filter.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "sitelinks"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/textvalues",
+          "name": "textvalues",
+          "description": "List of text values for the entity (name, aliases, description, generated type description).",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "textvalues"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/textsources",
+          "name": "textsources",
+          "description": "List of provenance tags aligned with textvalues (e.g. 'name', 'aliases', 'description').",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "textsources"
+            }
+          },
+          "repeated": true
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/is_superentity",
+          "name": "is_superentity",
+          "description": "Whether the entity is a super-entity (a parent grouping node).",
+          "dataType": "sc:Boolean",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "is_superentity"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/naturalentity",
+          "name": "naturalentity",
+          "description": "QID of the associated natural-language entity type.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "naturalentity"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/superentity",
+          "name": "superentity",
+          "description": "QID of the parent super-entity.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "superentity"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/graph",
+          "name": "graph",
+          "description": "Source knowledge graph (e.g. 'wikidata').",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "graph"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "entities/query_count",
+          "name": "query_count",
+          "description": "Number of queries generated for this entity.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "entitynet-entities"
+            },
+            "extract": {
+              "column": "query_count"
+            }
+          }
+        }
+      ]
+    }
+  ]
+}