| --- |
| license: mit |
| library_name: custom |
| pipeline_tag: text-classification |
| datasets: |
| - RKB109/support-routing-ml-20260811-dataset |
| tags: |
| - synthetic-data |
| - transparent-baseline |
| - applied-machine-learning |
| - text-classification |
| - zero-shot-classification |
| - sentence-similarity |
| - summarization |
| metrics: |
| - accuracy |
| --- |
| |
| # Applied ML Support Router Baseline Model |
|
|
| ## Model Description |
|
|
| This repository contains a small, transparent prototype model for |
| **Support operations need reproducible routing models that expose confidence and defer uncertain cases.** |
|
|
| The model combines per-label token weights with IDF-weighted evidence |
| retrieval. It was generated for reproducible architecture demonstrations and |
| does not call a hosted LLM. |
|
|
| ## Evaluation |
|
|
| - Held-out synthetic examples: 4 |
| - Accuracy: 1 |
| - Intended metrics: classification_accuracy, automation_coverage, escalation_precision |
| |
| ## Intended Use |
| |
| - Architecture prototyping |
| - CI and evaluation examples |
| - Local baseline comparisons |
| - Educational experimentation |
| |
| ## Hugging Face Task Coverage |
| |
| - `text-classification` |
| - `zero-shot-classification` |
| - `sentence-similarity` |
| - `summarization` |
| |
| ## Limitations and Risks |
| |
| Synthetic tickets do not represent every user population or language. Production training data needs consent and bias analysis. |
| |
| The dataset is synthetic and small. Do not use this model for consequential |
| decisions without representative data, expert review, and production-grade |
| evaluation. |
| |
| ## Reproducibility |
| |
| The linked GitHub repository includes `train.py`, the exact dataset split, |
| evaluation code, and the model JSON format. |
| |
