Wojood (وجود)

Named Entity Recognition

Nested and Flat NER, 21 entity types, and 31 entity subtypes. (Open Source)
Performance: Nested(89.42%) Flat(87.33%)

  • Models: Flat, Nested, and Fine-grain BERT models.

    Wojood Corpus: 560K tokens (MSA and dialect), manually annotated with 21 entity types, annotated with nested and flat entities (See Article). covers multiple domains (Media, History, Culture, Health, Finance, ICT, Law, Elections, Politics, Migration, Terrorism, social media).

    WojoodFine Corpus: Same as Wojood but extended with subtypes of entities (51 tags in total), (See Article).

    WojoodGaza Corpus: 60K tokens related to Israeli War on Gaza in domains (See Article).

    Tags and Guidelines:

    NORP (group of people) DATE ORDINAL
    OCC (occupation) TIME PERCENT
    ORG (organization) subtypes LANGUAGE QUANTITY
    GPE (geopolitical entity) subtypes WEBSITE UNIT
    LOC (geographical location) subtypes LAW MONEY
    FAC (facility: landmarks places) subtypes PRODUCT CURR (currency)
    ORG:
    GOV
    COM
    EDU
    ENT
    NONGOV
    MED
    REL
    SCI
    SPO
    ORG_FAC
    GPE:
    COUNTRY
    STATE-OR-PROVINCE
    TOWN
    NEIGHBORHOOD
    CAMP
    GPE_ORG
    SPORT
    LOC:
    CONTINENT
    CLUSTER
    ADDRESS
    BOUNDARY
    CELESTIAL
    WATER-BODY
    LAND-REGION-NATURAL
    REGION-GENERAL
    REGION-INTERNATIONAL
    FAC:
    PLANT
    AIRPORT
    BUILDING-OR-GROUNDS
    SUBAREA-FACILITY
    PATH

    Please email Prof. Jarrar (mjarrar AT birzeit.edu) for the annotation guidelines

  • SinaTools: NER module as python library.

    GitHub: training source code + sample data (~35K tokens).

    Hugging Face: fine-tuned BERT model using Wojood.

    Wojood Corpus (Corpus only)

    WojoodGaza Corpus (Corpus only)

    WojoodFine Corpus (Corpus only)

  • Evaluation of SinaTools NER module
    We benchmarked SinaTools using three datasets: (Wojood Test-set) and two additional out-of-domain datasets (WojoodGaza and Politics). We evaluated three NER scenarios: 3 tags (PERS, LOC and ORG), 21 Flat tags, and 21 Nested tags. See article [1] for the details.
    Wojood Benchmarks
    • WojoodNER-2024, The first Arabic Named Entity Recognition (NER) Shared Task.
    Page - Article

    WojoodNER-2023, The first Arabic Named Entity Recognition (NER) Shared Task.
    Page - Article

  • Mustafa Jarrar, Nagham Hamad, Mohammed Khalilia, Bashar Talafha, AbdelRahim Elmadany, Muhammad Abdul-Mageed: WojoodNER 2024: The Second Arabic Named Entity Recognition Shared Task. In Proceedings of the Second Arabic Natural Language Processing Conference (ArabicNLP 2024), Bangkok, Thailand. Association for Computational Linguistics.


    Mustafa Jarrar, Muhammad Abdul-Mageed, Mohammed Khalilia, Bashar Talafha, AbdelRahim El-madany, Nagham Hamad, Alaa’ Omar: WojoodNER 2023: The First Arabic Named Entity Recognition Shared Task. In Proceedings of the 1st Arabic Natural Language Processing Conference (Arabic- NLP), Part of the EMNLP 2023. ACL.


    Haneen Liqreina, Mustafa Jarrar, Mohammed Khalilia, Ahmed Oumar El-Shangiti, Muhammad AbdulMageed:Arabic Fine-Grained Entity Recognition.  In Proceedings of the 1st Arabic Natural Language Processing Conference (ArabicNLP), Part of the EMNLP 2023. ACL.


    Mustafa Jarrar, Mohammed Khalilia, Sana Ghanem: Wojood: Nested Arabic Named Entity Corpus and Recognition using BERT. In Proceedings of the International Conference on Language Resources and Evaluation (LREC 2022), Marseille, France. 2022