| --- |
| license: mit |
| language: en |
| tags: |
| - text-generation |
| - chatbot |
| - t5 |
| pipeline_tag: text2text-generation |
| --- |
| |
| # T5 Chatbot |
|
|
| A fine-tuned T5 model for conversational FAQ-style responses. Given a user question, the model generates a natural-language answer, making it suitable for lightweight chatbot and Q&A applications. |
|
|
| ## Model description |
|
|
| This model is a fine-tuned version of **[t5-small / t5-base]** (Google's T5 text-to-text transformer), adapted for conversational question-answering. It takes a user query as input and generates a relevant text response, framed as a text-to-text generation task. |
|
|
| ## Intended uses & limitations |
|
|
| **Intended uses:** |
| - FAQ-style chatbots for websites, apps, or customer support |
| - Educational/demo projects exploring conversational AI with T5 |
| - Quick prototyping of Q&A systems |
|
|
| **Limitations:** |
| - Trained on a limited dataset, so responses may be generic or repetitive outside the training domain |
| - Does not maintain multi-turn conversational context (treats each query independently) |
| - English only |
| - Not suitable for safety-critical or factual/medical/legal advice use cases |
| - May occasionally produce inaccurate or nonsensical answers (hallucination risk common to generative models) |
|
|
| ## Training data |
|
|
| The model was fine-tuned on a **[custom FAQ dataset / dataset name, e.g. "a collection of customer support Q&A pairs"]**. **[Add: dataset size, source, and any preprocessing steps if known.]** |
|
|
| ## How to use |
|
|
| ```python |
| from transformers import pipeline |
| |
| pipe = pipeline("text2text-generation", model="UMAR798/t5-chatbot") |
| response = pipe("What are your business hours?") |
| print(response) |
| ``` |
|
|
| ## Training procedure |
|
|
| - Base model: **[t5-small / t5-base]** |
| - Epochs: **[e.g. 3]** |
| - Batch size: **[e.g. 8]** |
| - Learning rate: **[e.g. 5e-5]** |
|
|
| ## Author |
|
|
| Developed by [Muhammad Umar Farooq](https://huggingface.co/UMAR798) — [LinkedIn](https://www.linkedin.com/in/muhammad-umar-farooq-6964a430b) |
