| --- |
| license: openrail |
| language: |
| - it |
| tags: |
| - italian |
| - dataset |
| - user-prompts |
| - chatbot |
| - Emma-5 |
| - anonymized |
| - classification |
| - conversational-ai |
| --- |
| |
| # Emma-5 User Prompts Dataset |
|
|
| This repository provides a **small public sample** of a larger dataset containing over **44,600 unique, real-world, anonymized user prompts** collected during the first launch phase of the **Emma** chatbot, in which users interacted with the **Emma-5** model from **June 19 to June 26, 2026**. |
|
|
| The original dataset contains prompts generated by real users and subsequently anonymized and classified into **21 categories**. Only a small sample, consisting of a few dozen prompts, is publicly available in this repository. |
|
|
| The complete dataset is not included in this repository. |
|
|
| ## Dataset Description |
|
|
| The original dataset consists of: |
|
|
| - **44,600+ unique prompts** |
| - Real prompts generated by users interacting with Emma-5 |
| - Data collected between **June 19 and June 26, 2026** |
| - **21 prompt categories** |
| - Fully anonymized content |
|
|
| The sample published in this repository is intended to provide a representative preview of the dataset's structure and content without making the complete dataset publicly available. |
|
|
| ## Intended Use |
|
|
| The public sample may be used for research, experimentation, dataset inspection, prompt classification, natural language processing, and conversational AI development. |
|
|
| The complete dataset may additionally be used for training, fine-tuning, and evaluation of machine learning and artificial intelligence models, subject to the applicable license and terms of use. |
|
|
| ## Privacy and Anonymization |
|
|
| The prompts in the original dataset were processed to remove or anonymize personal information before being included in the dataset. |
|
|
| The public sample available in this repository contains anonymized prompts only. |
|
|
| Users' identities are not included in the dataset. The dataset must not be used to attempt to identify, infer, or reconstruct the identity of any individual user. |
|
|
| ## Data Collection |
|
|
| The original dataset was collected during the initial launch phase of the Emma chatbot, between **June 19 and June 26, 2026**. |
|
|
| During this period, users interacted directly with the **Emma-5** model through the Emma chatbot. The resulting prompts were subsequently collected, anonymized, and classified into 21 categories. |
|
|
| ## Public Sample |
|
|
| Only a **small sample of a few dozen prompts** from the original dataset is publicly available through this repository. |
|
|
| The public sample does not contain the complete dataset and should not be considered representative of the full distribution of prompts across the original dataset. |
|
|
| ## License and Terms of Use |
|
|
| By downloading and using the publicly available sample, you agree to comply with the **OpenRAIL** license and the additional terms described below. |
|
|
| The publicly available sample may be used for research, experimentation, evaluation, and development purposes, subject to the applicable license terms. |
|
|
| Redistribution, resale, sublicensing, or sharing of the dataset or any substantial portion of it to third parties is prohibited. |
|
|
| Any model, technical documentation, model card, or public description of a system trained using this dataset must include the following attribution: |
|
|
| > "Trained using datasets provided by Egomnia S.p.A." |
|
|
| ## Complete Dataset |
|
|
| The complete dataset containing over **44,600 unique prompts** is not publicly available through this repository. |
|
|
| For information regarding access to the complete dataset and Egomnia's AI datasets, visit: |
|
|
| https://ai.egomnia.com/store |
|
|
| ## About Egomnia |
|
|
| **Egomnia S.p.A.** is an Italian innovative SME and AI research laboratory focused on artificial intelligence, software development, and Italian-language AI technologies. |
|
|
| Egomnia develops AI systems and datasets with a focus on Italian-language applications and real-world use cases. |
