Spaces:
Running on Zero
Running on Zero
| title: normaere | |
| emoji: 📜 | |
| colorFrom: blue | |
| colorTo: yellow | |
| sdk: gradio | |
| hardware: zero-gpu | |
| short_description: MHG Text Normalizer | |
| # normaere | |
| Transformer-based seq2seq-normalizer for Middle High German (MHG) text. | |
| Enter or upload MHG text and receive normalized output with line-by-line comparison. | |
| ## How It Works | |
| The model was fine-tuned on `google/mt5-base` (580M parameters) using the Referenzkorpus Mittelhochdeutsch (ReM). | |
| ## Features | |
| - Normalization of MHG texts according to the standards of ReM | |
| - Supported input: | |
| - All High German Dialects, ca. 1050–1500 | |
| - Diplomatic and modernized transcriptions, already normalized text (for standartization) | |
| - Solves common abbreviation | |
| - Splits off and expands common pro- and enclitics | |
| - Postprocessing to suit different editorial preferences | |
| ## Model | |
| - Base: `google/mt5-base` | |
| - Format: FP16 (reduced size, minimal quality loss) | |
| - Max sequence length: 512 tokens | |
| - Supports sliding-window inference for longer texts | |
| ## References | |
| - mT5: [Xue et al., 2021](https://arxiv.org/abs/2010.11934) | |
| - Training corpus: Roussel, Adam; Klein, Thomas; Dipper, Stefanie; Wegera, Klaus-Peter; Wich-Reif, Claudia (2024). Referenzkorpus Mittelhochdeutsch (1050–1350), Version 2.1, [https://www.linguistics.ruhr-uni-bochum.de/rem/](https://www.linguistics.ruhr-uni-bochum.de/rem/). ISLRN 937-948-254-174-0. |