dish OCR correction 

 

<!-- Provide a quick summary of what the model is/does. --> 

 

This model corrects OCR errors in Swedish text. 

utch OCR Correction  

 

This model is a finetuned byT5 model that corrects OCR mistakes found in dutch sentences. The [google/byt5-base](https://huggingface.co/google/byt5-base) model is finetuned on the dutch section of the [OSCAR](https://huggingface.co/datasets/oscar) dataset.  

 

 

for OCR Correction 

 

This model is a fine-tuned version of the [byt5-base](https://huggingface.co/google/byt5-base) for OCR Correction. ByT5 was 

introduced in [this paper](https://arxiv.org/abs/2105.13626) and the idea and code for fine-tuning the model for OCR Correction was taken from [here](https://blog.ml6.eu/ocr-correction-with-byt5-5994d1217c07). 

 

the correction of badly digitized texts, as part of the **Bad Data Toolbox**. 

 

OCROnos models are versatile tools supporting the correction of OCR errors, wrong word cut/merge and overall broken text structures. The training data includes a highly diverse set of ocrized texts in multiple languages from PleIAs open pre-training corpus, drawn from cultural heritage sources (Common Corpus) and financial and administrative documents in open data (Finance Commons). 

 

This release currently features a model based on llama-3-8b that has been the most tested to date. The model was trained using HPC resources from GENCI–IDRIS (Grant 2023-AD011014736) on Jean-Zay. Future release will focus on smaller internal models that provides a better ratio of generation cost/quality. 

-OCR Correction of Historical Newspapers](https://aclanthology.org/2024.lt4hala-1.14/) and designed to correct OCR text. [BART-base](https://huggingface.co/facebook/bart-base) is fine-tuned for post-OCR correction of historical English, using [BLN600](https://aclanthology.org/2024.lrec-main.219/), a parallel corpus of 19th century newspaper machine/human transcription. 

 

## Usage 

```python 

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline 

-OCR Correction of Historical Newspapers](https://aclanthology.org/2024.lt4hala-1.14/) and designed to correct OCR text. [BART-large](https://huggingface.co/facebook/bart-large) is fine-tuned for post-OCR correction of historical English, using [BLN600](https://aclanthology.org/2024.lrec-main.219/), a parallel corpus of 19th century newspaper machine/human transcription. 

 

## Usage 

```python 

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline 

-OCR Correction of Historical Newspapers](https://aclanthology.org/2024.lt4hala-1.14/) and designed to correct OCR text. [Llama 2 7B](https://huggingface.co/meta-llama/Llama-2-7b-hf) is instruction-tuned for post-OCR correction of historical English, using [BLN600](https://aclanthology.org/2024.lrec-main.219/), a parallel corpus of 19th century newspaper machine/human transcription. 

 

## Usage 

```python 

from peft import AutoPeftModelForCausalLM 

-OCR Correction of Historical Newspapers](https://aclanthology.org/2024.lt4hala-1.14/) and designed to correct OCR text. [Llama 2 13B](https://huggingface.co/meta-llama/Llama-2-13b-hf) is instruction-tuned for post-OCR correction of historical English, using [BLN600](https://aclanthology.org/2024.lrec-main.219/), a parallel corpus of 19th century newspaper machine/human transcription. 

 

## Usage 

```python 

from peft import AutoPeftModelForCausalLM 

### OCR correction 

 

```python 

text = """Otra vez, Don Iuan, me dad, 

y otras mil vezes los braços. 

# OCR postcorrection task 1 

 

This is a BertForTokenClassification model that predicts whether a token is an OCR  

mistake or not. It is based on [bert-base-multilingual-cased](https://huggingface.co/bert-base-multilingual-cased)  

and finetuned on the dataset of the  

lian OCR Error Correction Sequence-to-Sequence Model 

 

## Model Details 

This model represents the first version of an experimental sequence-to-sequence architecture designed specifically for the Italian language. It aims to correct approximately 93% of the errors generated by low-quality Optical Character Recognition (OCR) systems, which tend to perform poorly on Italian text. By taking raw, OCR-scanned text as input, the model outputs the corrected version of the text, significantly reducing errors and improving readability and accuracy. 

 

# ocr_correction 

 

This model was trained from scratch on the None dataset. 

 

## Model description 

rect OCR mistakes. The max_length was set to 350.  

 

## Performance  

``` 

SacreBLEU eval dataset: 10.83  

the correction of OCR errors and other digitization artifact. 

 

## Use 

Segmentext support the following text segmentation: 

* **Text** 

rror correction - error correction - correction | 40 | 57_gec_grammatical error_grammatical error correction_error correction |  

| 58 | intent - intent detection - slot - slot filling - filling | 40 | 58_intent_intent detection_slot_slot filling |  

| 59 | temporal - events - temporal relations - expressions - temporal relation | 39 | 59_temporal_events_temporal relations_expressions |  

| 60 | adaptation - domain - domain adaptation - indomain - translation | 37 | 60_adaptation_domain_domain adaptation_indomain |  

| 61 | stance - stance detection - detection - tweets - veracity | 37 | 61_stance_stance detection_detection_tweets |  

ling Correction v1 

 

This model is a version of [google/canine-c](https://huggingface.co/openai/whisper-small) fine-tuned to fix corrupted texts. 

It was trained on a mixture of two parallel datasets in the Bashkir language: 

- sentences post-edited by humans after OCR 

mmar correction on an expanded version of the [JFLEG](https://paperswithcode.com/dataset/jfleg) dataset. 

 

usage in Python (after `pip install transformers`): 

 

```python 

mmar correction on an expanded version of the [JFLEG](https://paperswithcode.com/dataset/jfleg) dataset. Check out a [demo notebook on Colab here](https://colab.research.google.com/gist/pszemraj/91abb08aa99a14d9fdc59e851e8aed66/demo-for-grammar-synthesis-base.ipynb). 

 

usage in Python (after `pip install transformers`): 

 

```python 

mmar correction on an expanded version of the [JFLEG](https://paperswithcode.com/dataset/jfleg) dataset. 

 

usage in Python (after `pip install transformers`): 

 

```python 

mmar correction on an expanded version of the [JFLEG](https://paperswithcode.com/dataset/jfleg) dataset. [Demo](https://huggingface.co/spaces/pszemraj/FLAN-grammar-correction) on HF spaces. 

 

## Example