Symptom to Diagnosis
A textual data set focused on the recognition of diagnoses from natural language symptom descriptions, useful for supervised medical classification.
1,065 examples in JSONL, input/output fields in English, divided into train/test
Apache 2.0
Description
The dataset Symptom to Diagnosis contains 1,065 “symptoms → diagnosis” pairs, written in natural English simulating patient descriptions. The texts were generated or cleaned using an LLM for greater linguistic consistency. Each entry has a field input_text (symptoms) and output_text (diagnosis).
What is this dataset for?
- Form a supervised medical text classification model (symptom → diagnosis)
- Creating decision support systems for general medicine
- Analyze the language used by patients in describing medical conditions
Can it be enriched or improved?
Yes. It is possible to add linguistic variants, multilingual translations or to annotate cases by severity or chronology of symptoms. Labelling can also be refined according to clinical sub-categories.
🔎 In summary
🧠 Recommended for
- Medical NLP classification projects
- Triage assistants
- Health AI student training
🔧 Compatible tools
- Scikit-learn
- Hugging Face Transformers
- FastText
- SpacY
💡 Tip
For better robustness, add paraphrases and synonyms in symptomatic descriptions.
Frequently Asked Questions
Can this dataset be used to train an automated medical diagnostic model?
Yes, that is precisely its objective. It makes it possible to train an NLP model to match textual symptoms to a probable diagnosis.
Is this dataset suitable for multilingual use?
Currently it is only in English. However, it can be translated and enriched to cover other languages or clinical contexts.
What is the granularity of the diagnoses in this dataset?
The dataset covers 22 common diagnoses, each represented by forty examples, which allows for balanced and targeted learning.
