Mayssm/rafeeq-arabic-medical-dialect

Rafiq Arabic Medical Dialect Dataset

Dataset summary

Rafiq is an Arabic dataset for dialect-to-simple-Arabic normalization/translation of health-related expressions.

The primary task is to convert a colloquial Arabic expression into simpler Arabic while preserving its meaning and avoiding added diagnoses or symptoms.

medical_department is provided as an optional auxiliary text-classification/routing label. It is not the primary task and must not be interpreted as a diagnosis, clinical triage rule, or medical recommendation.

Dataset structure

Related paraphrases sharing the same group_id are kept in one split to reduce train/evaluation leakage.

Fields

• —group_id: identifier for related linguistic/paraphrase examples.
• —dialect_text: original dialectal/colloquial Arabic expression.
• —simple_arabic: simplified Arabic rendering preserving the source meaning.
• —medical_department: optional auxiliary routing/classification label.

Medical department labels

• —أنف وأذن وحنجرة
• —باطنية
• —طب عام / طب الأسرة
• —طوارئ
• —عظام ومفاصل
• —نفسية

Tasks

Primary task: dialect normalization / translation

Input: dialect_text Target: simple_arabic

Secondary optional task: text classification

Input: dialect_text or simple_arabic Label: medical_department

Annotation policy for the auxiliary label

The classification labels were normalized for internal consistency:

• —Explicit injury/accident descriptions are labeled طوارئ.
• —A symptom may be labeled طوارئ when the text itself explicitly states severe/incapacitating intensity.
• —Otherwise the symptom-appropriate department represented by the dataset is used.
• —Nosebleed without stated trauma/severe danger markers is labeled أنف وأذن وحنجرة; trauma/severe emergency context may be labeled طوارئ.

These are dataset annotation conventions, not clinical triage guidelines.

Quality controls

For this release:

• —all 493 examples have complete required fields;
• —all rows have group_id;
• —train/validation/test splits are group-aware;
• —no group crosses multiple splits;
• —every medical-department class appears in every split;
• —no exact duplicate publication rows were detected;
• —confirmed typographical and normalization-direction errors were corrected;
• —all previously flagged language-review cases were resolved.

Two confirmed dialect decisions from the final review:

• —اندبجوا → انضربوا
• —أثور in the reviewed context → أقوم/أنهض

Intended uses

• —Arabic dialect normalization and translation research.
• —Dialect-aware NLP prototypes.
• —Training/evaluating text-to-text normalization models.
• —Optional exploratory medical-department classification experiments.

Limitations

• —The dataset is small and does not represent all Saudi or Arabic dialects.
• —Dialect coverage is not guaranteed to be geographically or demographically balanced.
• —Dialect expressions may be context-dependent.
• —The medical-department field is an auxiliary annotation and is not validated for clinical use.
• —Do not use this dataset for autonomous diagnosis, treatment recommendations, or clinical triage.

Provenance and creation

The dialect vocabulary in this dataset was collected and curated manually by the Rafiq team based on Arabic dialect expressions known to the team.

Example sentences were created using a mixed human/AI-assisted workflow:

• —some example sentences were written manually by team members;
• —some example sentences were generated with the assistance of generative AI;
• —paraphrase expansion and sentence variation were also performed with generative AI assistance.

The team manually reviewed the dataset during preparation to correct wording, preserve the intended dialect meaning, remove obvious inconsistencies, and reduce label noise.

The simplified-Arabic field was reviewed with the goal of preserving the meaning of the original dialect expression without adding a diagnosis or unsupported symptom.

The medical_department field is an auxiliary annotation intended for optional classification/routing experiments and is not a clinical diagnosis or medical recommendation.

AI-assisted content disclosure

This dataset contains AI-assisted synthetic/augmented text. Generative AI was used for part of the sentence creation and for paraphrase/data augmentation. The underlying dialect vocabulary itself was collected manually by the team.

Users should take this into account when evaluating linguistic diversity, naturalness, and real-world representativeness.

Privacy

No direct personal identifiers were found during the publication review. If any examples originated from real individuals or real clinical interactions, the publishing team should confirm that redistribution, consent, and de-identification requirements are satisfied.

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

You may share and adapt the dataset, including commercially, provided appropriate attribution is given and changes are indicated.

License: https://creativecommons.org/licenses/by/4.0/

This license applies only to content that the publishing team has the right to license.

Citation

Update the team/author field if needed before publication:

@dataset{rafeeq_arabic_medical_dialect_2026,
  author    = {Rafiq Team},
  title     = {Rafiq Arabic Medical Dialect Dataset},
  year      = {2026},
  publisher = {Hugging Face}
}