NADI2024-baseline

Model Card for NADI-2024-baseline

<!-- Provide a quick summary of what the model is/does. -->

A BERT-based model fine-tuned to perform single-label Arabic Dialect Identification (ADI). Instead of predicting the most probable dialect, the logits are used to generate multilabel predictions.

Model Description

<!-- Provide a longer summary of what this model is. --> <!-- - **Developed by:** Amr Keleg -->

• Model type: A Dialect Identification model fine-tuned on the training sets of: NADI2020,2021,2023 and MADAR 2018.
• Language(s) (NLP): Arabic.

<!--- **License:** [More Information Needed] -->

• Finetuned from model : MarBERTv2

Multilabel country-level Dialect Identification

#### Baseline I (Top 90%)

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

DIALECTS = ["Algeria",
    "Bahrain",
    "Egypt",
    "Iraq",
    "Jordan",
    "Kuwait",
    "Lebanon",
    "Libya",
    "Morocco",
    "Oman",
    "Palestine",
    "Qatar",
    "Saudi_Arabia",
    "Sudan",
    "Syria",
    "Tunisia",
    "UAE",
    "Yemen",
]
assert len(DIALECTS) == 18

MODEL_NAME = "AMR-KELEG/NADI2024-baseline"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)

def predict_top_p(text, P=0.9):
    """Predict the top dialects with an accumulative confidence of at least P."""
    assert P <= 1 and P >= 0

    logits = model(**tokenizer(text, return_tensors="pt")).logits
    probabilities = torch.softmax(logits, dim=1).flatten().tolist()
    topk_predictions = torch.topk(logits, 18).indices.flatten().tolist()

    predictions = [0 for _ in range(18)]
    total_prob = 0

    for i in range(18):
        total_prob += probabilities[topk_predictions[i]]
        predictions[topk_predictions[i]] = 1
        if total_prob >= P:
            break

    return [DIALECTS[i] for i, p in enumerate(predictions) if p == 1]

s1 = "كيفك يا زلمة"
s1_pred = predict_top_p(s1) # ['Jordan', 'Lebanon', 'Palestine', 'Syria']
print(s1, s1_pred)

s2 = "خليلي في مساج بريفي كيفاش الاتصال"
s2_pred = predict_top_p(s2) # ['Algeria', 'Tunisia']
print(s2, s2_pred)

Citation

If you find the model useful, please cite the following respective paper:

@inproceedings{abdul-mageed-etal-2024-nadi,
    title = "{NADI} 2024: The Fifth Nuanced {A}rabic Dialect Identification Shared Task",
    author = "Abdul-Mageed, Muhammad  and
      Keleg, Amr  and
      Elmadany, AbdelRahim  and
      Zhang, Chiyu  and
      Hamed, Injy  and
      Magdy, Walid  and
      Bouamor, Houda  and
      Habash, Nizar",
    editor = "Habash, Nizar  and
      Bouamor, Houda  and
      Eskander, Ramy  and
      Tomeh, Nadi  and
      Abu Farha, Ibrahim  and
      Abdelali, Ahmed  and
      Touileb, Samia  and
      Hamed, Injy  and
      Onaizan, Yaser  and
      Alhafni, Bashar  and
      Antoun, Wissam  and
      Khalifa, Salam  and
      Haddad, Hatem  and
      Zitouni, Imed  and
      AlKhamissi, Badr  and
      Almatham, Rawan  and
      Mrini, Khalil",
    booktitle = "Proceedings of The Second Arabic Natural Language Processing Conference",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.arabicnlp-1.79",
    pages = "709--728",
    abstract = "We describe the findings of the fifth Nuanced Arabic Dialect Identification Shared Task (NADI 2024). NADI{'}s objective is to help advance SoTA Arabic NLP by providing guidance, datasets, modeling opportunities, and standardized evaluation conditions that allow researchers to collaboratively compete on prespecified tasks. NADI 2024 targeted both dialect identification cast as a multi-label task (Subtask 1), identification of the Arabic level of dialectness (Subtask 2), and dialect-to-MSA machine translation (Subtask 3). A total of 51 unique teams registered for the shared task, of whom 12 teams have participated (with 76 valid submissions during the test phase). Among these, three teams participated in Subtask 1, three in Subtask 2, and eight in Subtask 3. The winning teams achieved 50.57 F1 on Subtask 1, 0.1403 RMSE for Subtask 2, and 20.44 BLEU in Subtask 3, respectively. Results show that Arabic dialect processing tasks such as dialect identification and machine translation remain challenging. We describe the methods employed by the participating teams and briefly offer an outlook for NADI.",
}