mteb 2.1.5__py3-none-any.whl → 2.1.7__py3-none-any.whl

mteb/tasks/clustering/nld/dutch_news_articles_clustering_p2p.py

@@ -28,6 +28,9 @@ class DutchNewsArticlesClusteringP2P(AbsTaskClustering):
         dialect=[],
         sample_creation="found",
         bibtex_citation="",
+        prompt={
+            "query": "Identificeer de hoofdcategorie van nieuwsartikelen op basis van de titels en de inhoud"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/clustering/nld/dutch_news_articles_clustering_s2s.py

@@ -28,6 +28,9 @@ class DutchNewsArticlesClusteringS2S(AbsTaskClustering):
         dialect=[],
         sample_creation="found",
         bibtex_citation="",
+        prompt={
+            "query": "Identificeer de hoofdcategorie van nieuwsartikelen op basis van de titels"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/clustering/nld/iconclass_clustering_s2s.py

@@ -38,6 +38,9 @@ class IconclassClusteringS2S(AbsTaskClustering):
   year = {2023},
 }
 """,
+        prompt={
+            "query": "Identificeer het onderwerp of thema van kunstwerken op basis van de titels"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/clustering/nld/open_tender_clustering_p2p.py

@@ -38,6 +38,9 @@ class OpenTenderClusteringP2P(AbsTaskClustering):
   year = {2025},
 }
 """,
+        prompt={
+            "query": "Identificeer de hoofdcategorie van aanbestedingen op basis van de titels en beschrijvingen"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/clustering/nld/open_tender_clustering_s2s.py

@@ -38,4 +38,7 @@ class OpenTenderClusteringS2S(AbsTaskClustering):
   year = {2025},
 }
 """,
+        prompt={
+            "query": "Identificeer de hoofdcategorie van aanbestedingen op basis van de titels"
+        },
     )

mteb/tasks/clustering/nld/vabb_clustering_p2p.py

@@ -39,6 +39,9 @@ class VABBClusteringP2P(AbsTaskClustering):
   year = {2024},
 }
 """,
+        prompt={
+            "query": "Identificeer de hoofdcategorie van wetenschappelijke artikelen op basis van de titels en abstracts"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/clustering/nld/vabb_clustering_s2s.py

@@ -39,6 +39,9 @@ class VABBClusteringS2S(AbsTaskClustering):
   year = {2024},
 }
 """,
+        prompt={
+            "query": "Identificeer de hoofdcategorie van wetenschappelijke artikelen op basis van de titels"
+        },
     )
 
     def dataset_transform(self):

mteb/tasks/multilabel_classification/nld/covid_disinformation_nl_multi_label_classification.py

@@ -61,6 +61,9 @@ Yih, Scott Wen-tau},
   year = {2021},
 }
 """,
+        prompt={
+            "query": "Classificeer COVID-19-gerelateerde sociale media-berichten in alle toepasselijke desinformatiecategorieën"
+        },
     )
 
     def dataset_transform(self) -> None:

mteb/tasks/multilabel_classification/nld/vabb_multi_label_classification.py

@@ -41,4 +41,7 @@ class VABBMultiLabelClassification(AbsTaskMultilabelClassification):
   year = {2024},
 }
 """,
+        prompt={
+            "query": "Classificeer de onderwerpen van een wetenschappelijk artikel op basis van de abstract"
+        },
     )

mteb/tasks/pair_classification/nld/sick_nl_pair_classification.py

@@ -33,4 +33,7 @@ class SICKNLPairClassification(AbsTaskPairClassification):
   year = {2021},
 }
 """,
+        prompt={
+            "query": "Zoek tekst die semantisch vergelijkbaar is met de gegeven tekst."
+        },
     )

mteb/tasks/pair_classification/nld/xlwic_nl_pair_classification.py

@@ -38,4 +38,7 @@ class XLWICNLPairClassification(AbsTaskPairClassification):
   year = {2020},
 }
 """,
+        prompt={
+            "query": "Zoek tekst die semantisch vergelijkbaar is met de gegeven tekst."
+        },
     )

mteb/tasks/retrieval/nld/__init__.py

@@ -1,4 +1,4 @@
-from .argu_ana_nl_retrieval import ArguAnaNL
+from .argu_ana_nl_retrieval import ArguAnaNL, ArguAnaNLv2
 from .bbsard_nl_retrieval import BBSARDNLRetrieval
 from .climate_fevernl_retrieval import ClimateFEVERNL
 from .cqa_dupstack_android_nl_retrieval import CQADupstackAndroidNLRetrieval
@@ -20,12 +20,12 @@ from .fi_qa2018_nl_retrieval import FiQA2018NL
 from .hotpot_qanl_retrieval import HotpotQANL
 from .legal_qa_nl_retrieval import LegalQANLRetrieval
 from .mmarconl_retrieval import MMMARCONL
-from .nf_corpus_nl_retrieval import NFCorpusNL
+from .nf_corpus_nl_retrieval import NFCorpusNL, NFCorpusNLv2
 from .nqnl_retrieval import NQNL
 from .open_tender_retrieval import OpenTenderRetrieval
 from .quora_nl_retrieval import QuoraNLRetrieval
-from .sci_fact_nl_retrieval import SciFactNL
-from .scidocsnl_retrieval import SCIDOCSNL
+from .sci_fact_nl_retrieval import SciFactNL, SciFactNLv2
+from .scidocsnl_retrieval import SCIDOCSNL, SCIDOCSNLv2
 from .touche2020_nl_retrieval import Touche2020NL
 from .treccovidnl_retrieval import TRECCOVIDNL
 from .vabb_retrieval import VABBRetrieval
@@ -37,6 +37,7 @@ __all__ = [
     "SCIDOCSNL",
     "TRECCOVIDNL",
     "ArguAnaNL",
+    "ArguAnaNLv2",
     "BBSARDNLRetrieval",
     "CQADupstackAndroidNLRetrieval",
     "CQADupstackEnglishNLRetrieval",
@@ -57,9 +58,12 @@ __all__ = [
     "HotpotQANL",
     "LegalQANLRetrieval",
     "NFCorpusNL",
+    "NFCorpusNLv2",
     "OpenTenderRetrieval",
     "QuoraNLRetrieval",
+    "SCIDOCSNLv2",
     "SciFactNL",
+    "SciFactNLv2",
     "Touche2020NL",
     "VABBRetrieval",
 ]

mteb/tasks/retrieval/nld/argu_ana_nl_retrieval.py

@@ -1,33 +1,26 @@
 from mteb.abstasks.retrieval import AbsTaskRetrieval
 from mteb.abstasks.task_metadata import TaskMetadata
 
-
-class ArguAnaNL(AbsTaskRetrieval):
-    ignore_identical_ids = True
-
-    metadata = TaskMetadata(
-        name="ArguAna-NL",
-        description="ArguAna involves the task of retrieval of the best counterargument to an argument. ArguAna-NL is "
-        "a Dutch translation.",
-        reference="https://huggingface.co/datasets/clips/beir-nl-arguana",
-        dataset={
-            "path": "clips/beir-nl-arguana",
-            "revision": "4cd085d148fe2cac923bb7758d6ef585926170ba",
-        },
-        type="Retrieval",
-        category="t2t",
-        modalities=["text"],
-        eval_splits=["test"],
-        eval_langs=["nld-Latn"],
-        main_score="ndcg_at_10",
-        date=("2016-03-01", "2016-03-01"),  # best guess: based on publication date
-        domains=["Written", "Non-fiction"],
-        task_subtypes=[],
-        license="cc-by-sa-4.0",
-        annotations_creators="derived",
-        dialect=[],
-        sample_creation="machine-translated and verified",  # manually checked a small subset
-        bibtex_citation=r"""
+_argu_ana_nl_metadata = dict(
+    reference="https://huggingface.co/datasets/clips/beir-nl-arguana",
+    dataset={
+        "path": "clips/beir-nl-arguana",
+        "revision": "4cd085d148fe2cac923bb7758d6ef585926170ba",
+    },
+    type="Retrieval",
+    category="t2t",
+    modalities=["text"],
+    eval_splits=["test"],
+    eval_langs=["nld-Latn"],
+    main_score="ndcg_at_10",
+    date=("2016-03-01", "2016-03-01"),  # best guess: based on publication date
+    domains=["Written", "Non-fiction"],
+    task_subtypes=[],
+    license="cc-by-sa-4.0",
+    annotations_creators="derived",
+    dialect=[],
+    sample_creation="machine-translated and verified",  # manually checked a small subset
+    bibtex_citation=r"""
 @misc{banar2024beirnlzeroshotinformationretrieval,
   archiveprefix = {arXiv},
   author = {Nikolay Banar and Ehsan Lotfi and Walter Daelemans},
@@ -38,5 +31,31 @@ class ArguAnaNL(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+)
+
+
+class ArguAnaNL(AbsTaskRetrieval):
+    ignore_identical_ids = True
+
+    metadata = TaskMetadata(
+        name="ArguAna-NL",
+        description="ArguAna involves the task of retrieval of the best counterargument to an argument. ArguAna-NL is "
+        "a Dutch translation.",
         adapted_from=["ArguAna"],
+        **_argu_ana_nl_metadata,
+    )
+
+
+class ArguAnaNLv2(AbsTaskRetrieval):
+    ignore_identical_ids = True
+
+    metadata = TaskMetadata(
+        name="ArguAna-NL.v2",
+        description="ArguAna involves the task of retrieval of the best counterargument to an argument. ArguAna-NL is "
+        "a Dutch translation. This version adds a Dutch prompt to the dataset.",
+        prompt={
+            "query": "Gegeven een bewering, vind documenten die de bewering weerleggen"
+        },
+        adapted_from=["ArguAna-NL"],
+        **_argu_ana_nl_metadata,
     )

mteb/tasks/retrieval/nld/bbsard_nl_retrieval.py

@@ -38,4 +38,7 @@ class BBSARDNLRetrieval(AbsTaskRetrieval):
   year = {2025},
 }
 """,
+        prompt={
+            "query": "Gegeven een juridische vraag, haal documenten op die kunnen helpen bij het beantwoorden van de vraag"
+        },
     )

mteb/tasks/retrieval/nld/dutch_news_articles_retrieval.py

@@ -27,4 +27,7 @@ class DutchNewsArticlesRetrieval(AbsTaskRetrieval):
         dialect=[],
         sample_creation="found",
         bibtex_citation="",
+        prompt={
+            "query": "Gegeven een titel, haal het nieuwsartikel op dat het beste bij de titel past"
+        },
     )

mteb/tasks/retrieval/nld/legal_qa_nl_retrieval.py

@@ -36,4 +36,7 @@ class LegalQANLRetrieval(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+        prompt={
+            "query": "Gegeven een juridische vraag, haal documenten op die kunnen helpen bij het beantwoorden van de vraag"
+        },
     )

mteb/tasks/retrieval/nld/nf_corpus_nl_retrieval.py

@@ -1,31 +1,26 @@
 from mteb.abstasks.retrieval import AbsTaskRetrieval
 from mteb.abstasks.task_metadata import TaskMetadata
 
-
-class NFCorpusNL(AbsTaskRetrieval):
-    metadata = TaskMetadata(
-        name="NFCorpus-NL",
-        dataset={
-            "path": "clips/beir-nl-nfcorpus",
-            "revision": "942953e674fd0f619ff89897abb806dc3df5dd39",
-        },
-        description="NFCorpus: A Full-Text Learning to Rank Dataset for Medical Information Retrieval. NFCorpus-NL is "
-        "a Dutch translation.",
-        reference="https://huggingface.co/datasets/clips/beir-nl-nfcorpus",
-        type="Retrieval",
-        category="t2t",
-        modalities=["text"],
-        eval_splits=["test"],
-        eval_langs=["nld-Latn"],
-        main_score="ndcg_at_10",
-        date=("2016-03-01", "2016-03-01"),  # best guess: based on publication date
-        domains=["Medical", "Academic", "Written"],
-        task_subtypes=[],
-        license="cc-by-4.0",
-        annotations_creators="derived",
-        dialect=[],
-        sample_creation="machine-translated and verified",  # manually checked a small subset
-        bibtex_citation=r"""
+_nf_corpus_metadata = dict(
+    dataset={
+        "path": "clips/beir-nl-nfcorpus",
+        "revision": "942953e674fd0f619ff89897abb806dc3df5dd39",
+    },
+    reference="https://huggingface.co/datasets/clips/beir-nl-nfcorpus",
+    type="Retrieval",
+    category="t2t",
+    modalities=["text"],
+    eval_splits=["test"],
+    eval_langs=["nld-Latn"],
+    main_score="ndcg_at_10",
+    date=("2016-03-01", "2016-03-01"),  # best guess: based on publication date
+    domains=["Medical", "Academic", "Written"],
+    task_subtypes=[],
+    license="cc-by-4.0",
+    annotations_creators="derived",
+    dialect=[],
+    sample_creation="machine-translated and verified",  # manually checked a small subset
+    bibtex_citation=r"""
 @misc{banar2024beirnlzeroshotinformationretrieval,
   archiveprefix = {arXiv},
   author = {Nikolay Banar and Ehsan Lotfi and Walter Daelemans},
@@ -36,5 +31,27 @@ class NFCorpusNL(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+)
+
+
+class NFCorpusNL(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="NFCorpus-NL",
+        description="NFCorpus: A Full-Text Learning to Rank Dataset for Medical Information Retrieval. NFCorpus-NL is "
+        "a Dutch translation.",
         adapted_from=["NFCorpus"],
+        **_nf_corpus_metadata,
+    )
+
+
+class NFCorpusNLv2(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="NFCorpus-NL.v2",
+        description="NFCorpus: A Full-Text Learning to Rank Dataset for Medical Information Retrieval. NFCorpus-NL is "
+        "a Dutch translation. This version adds a Dutch prompt to the dataset.",
+        adapted_from=["NFCorpus-NL"],
+        prompt={
+            "query": "Gegeven een vraag, haal relevante documenten op die de vraag het beste beantwoorden"
+        },
+        **_nf_corpus_metadata,
     )

mteb/tasks/retrieval/nld/open_tender_retrieval.py

@@ -35,4 +35,7 @@ class OpenTenderRetrieval(AbsTaskRetrieval):
   year = {2025},
 }
 """,
+        prompt={
+            "query": "Gegeven een titel, haal de aanbestedingsbeschrijving op die het beste bij de titel past"
+        },
     )

mteb/tasks/retrieval/nld/sci_fact_nl_retrieval.py

@@ -1,30 +1,26 @@
 from mteb.abstasks.retrieval import AbsTaskRetrieval
 from mteb.abstasks.task_metadata import TaskMetadata
 
-
-class SciFactNL(AbsTaskRetrieval):
-    metadata = TaskMetadata(
-        name="SciFact-NL",
-        dataset={
-            "path": "clips/beir-nl-scifact",
-            "revision": "856d8dfc294b138856bbf3042450e3782321e44e",
-        },
-        description="SciFactNL verifies scientific claims in Dutch using evidence from the research literature containing scientific paper abstracts.",
-        reference="https://huggingface.co/datasets/clips/beir-nl-scifact",
-        type="Retrieval",
-        category="t2t",
-        modalities=["text"],
-        eval_splits=["test"],
-        eval_langs=["nld-Latn"],
-        main_score="ndcg_at_10",
-        date=("2020-05-01", "2020-05-01"),  # best guess: based on submission date
-        domains=["Academic", "Medical", "Written"],
-        task_subtypes=[],
-        license="cc-by-4.0",
-        annotations_creators="derived",
-        dialect=[],
-        sample_creation="machine-translated and verified",  # manually checked a small subset
-        bibtex_citation=r"""
+_sci_fact_nl_metadata = dict(
+    dataset={
+        "path": "clips/beir-nl-scifact",
+        "revision": "856d8dfc294b138856bbf3042450e3782321e44e",
+    },
+    reference="https://huggingface.co/datasets/clips/beir-nl-scifact",
+    type="Retrieval",
+    category="t2t",
+    modalities=["text"],
+    eval_splits=["test"],
+    eval_langs=["nld-Latn"],
+    main_score="ndcg_at_10",
+    date=("2020-05-01", "2020-05-01"),  # best guess: based on submission date
+    domains=["Academic", "Medical", "Written"],
+    task_subtypes=[],
+    license="cc-by-4.0",
+    annotations_creators="derived",
+    dialect=[],
+    sample_creation="machine-translated and verified",  # manually checked a small subset
+    bibtex_citation=r"""
 @misc{banar2024beirnlzeroshotinformationretrieval,
   archiveprefix = {arXiv},
   author = {Nikolay Banar and Ehsan Lotfi and Walter Daelemans},
@@ -35,5 +31,27 @@ class SciFactNL(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+)
+
+
+class SciFactNL(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="SciFact-NL",
+        description="SciFactNL verifies scientific claims in Dutch using evidence from the research literature "
+        "containing scientific paper abstracts.",
         adapted_from=["SciFact"],
+        **_sci_fact_nl_metadata,
+    )
+
+
+class SciFactNLv2(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="SciFact-NL.v2",
+        description="SciFactNL verifies scientific claims in Dutch using evidence from the research literature "
+        "containing scientific paper abstracts. This version adds a Dutch prompt to the dataset.",
+        adapted_from=["SciFact-NL"],
+        prompt={
+            "query": "Given a scientific claim, retrieve documents that support or refute the claim"
+        },
+        **_sci_fact_nl_metadata,
     )

mteb/tasks/retrieval/nld/scidocsnl_retrieval.py

@@ -1,33 +1,26 @@
 from mteb.abstasks.retrieval import AbsTaskRetrieval
 from mteb.abstasks.task_metadata import TaskMetadata
 
-
-class SCIDOCSNL(AbsTaskRetrieval):
-    metadata = TaskMetadata(
-        name="SCIDOCS-NL",
-        dataset={
-            "path": "clips/beir-nl-scidocs",
-            "revision": "4e018aa220029f9d1bd5a31de3650e322e32ea38",
-        },
-        description=(
-            "SciDocs, a new evaluation benchmark consisting of seven document-level tasks ranging from citation"
-            + " prediction, to document classification and recommendation. SciDocs-NL is a Dutch translation."
-        ),
-        reference="https://huggingface.co/datasets/clips/beir-nl-scidocs",
-        type="Retrieval",
-        category="t2t",
-        modalities=["text"],
-        eval_splits=["test"],
-        eval_langs=["nld-Latn"],
-        main_score="ndcg_at_10",
-        date=("2020-05-01", "2020-05-01"),  # best guess: based on submission date
-        domains=["Academic", "Written", "Non-fiction"],
-        task_subtypes=[],
-        license="cc-by-sa-4.0",
-        annotations_creators="derived",
-        dialect=[],
-        sample_creation="machine-translated and verified",  # manually checked a small subset
-        bibtex_citation=r"""
+_scidocsnl_metadata = dict(
+    dataset={
+        "path": "clips/beir-nl-scidocs",
+        "revision": "4e018aa220029f9d1bd5a31de3650e322e32ea38",
+    },
+    reference="https://huggingface.co/datasets/clips/beir-nl-scidocs",
+    type="Retrieval",
+    category="t2t",
+    modalities=["text"],
+    eval_splits=["test"],
+    eval_langs=["nld-Latn"],
+    main_score="ndcg_at_10",
+    date=("2020-05-01", "2020-05-01"),  # best guess: based on submission date
+    domains=["Academic", "Written", "Non-fiction"],
+    task_subtypes=[],
+    license="cc-by-sa-4.0",
+    annotations_creators="derived",
+    dialect=[],
+    sample_creation="machine-translated and verified",  # manually checked a small subset
+    bibtex_citation=r"""
 @misc{banar2024beirnlzeroshotinformationretrieval,
   archiveprefix = {arXiv},
   author = {Nikolay Banar and Ehsan Lotfi and Walter Daelemans},
@@ -38,5 +31,29 @@ class SCIDOCSNL(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+)
+
+
+class SCIDOCSNL(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="SCIDOCS-NL",
+        description="SciDocs, a new evaluation benchmark consisting of seven document-level tasks ranging from "
+        "citation prediction, to document classification and recommendation. SciDocs-NL is a Dutch "
+        "translation.",
         adapted_from=["SCIDOCS"],
+        **_scidocsnl_metadata,
+    )
+
+
+class SCIDOCSNLv2(AbsTaskRetrieval):
+    metadata = TaskMetadata(
+        name="SCIDOCS-NL.v2",
+        description="SciDocs, a new evaluation benchmark consisting of seven document-level tasks ranging from "
+        "citation prediction, to document classification and recommendation. SciDocs-NL is a Dutch "
+        "translation. This version adds a Dutch prompt to the dataset.",
+        adapted_from=["SCIDOCS-NL"],
+        **_scidocsnl_metadata,
+        prompt={
+            "query": "Gegeven de titel van een wetenschappelijk artikel, haal de abstracts op van artikelen die door het gegeven artikel worden geciteerd"
+        },
     )

mteb/tasks/retrieval/nld/vabb_retrieval.py

@@ -38,4 +38,7 @@ class VABBRetrieval(AbsTaskRetrieval):
   year = {2024},
 }
 """,
+        prompt={
+            "query": "Gegeven een titel, haal de wetenschappelijke abstract op die het beste bij de titel past"
+        },
     )

mteb/tasks/sts/nld/sick_nl_sts.py

@@ -35,6 +35,7 @@ class SICKNLSTS(AbsTaskSTS):
   year = {2021},
 }
 """,
+        prompt={"query": "Haal semantisch vergelijkbare tekst op"},
     )
 
     min_score = 0

{mteb-2.1.5.dist-info → mteb-2.1.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mteb
-Version: 2.1.5
+Version: 2.1.7
 Summary: Massive Text Embedding Benchmark
 Author-email: MTEB Contributors <niklas@huggingface.co>, Kenneth Enevoldsen <kenneth.enevoldsen@cas.au.dk>, Nouamane Tazi <nouamane@huggingface.co>, Nils Reimers <info@nils-reimers.de>
 Maintainer-email: Kenneth Enevoldsen <kenneth.enevoldsen@cas.au.dk>, Roman Solomatin <risolomatin@gmail.com>, Isaac Chung <chungisaac1217@gmail.com>
@@ -16,7 +16,7 @@ Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Information Technology
 Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python
-Requires-Python: <3.14,>=3.10
+Requires-Python: <3.15,>=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: datasets>=2.19.0