Text Classification
Transformers
Safetensors
English
modernbert
legal
document-ai
page-classification
text-embeddings-inference
Instructions to use RayJackson30/clawbert-149 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use RayJackson30/clawbert-149 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="RayJackson30/clawbert-149")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForSequenceClassification tokenizer = AutoTokenizer.from_pretrained("RayJackson30/clawbert-149") model = AutoModelForSequenceClassification.from_pretrained("RayJackson30/clawbert-149", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -33,10 +33,10 @@ Exemplars (individual images in [docs/examples/](docs/examples/)):
|
|
| 33 |
|
| 34 |
## Purpose
|
| 35 |
|
| 36 |
-
1. **Routing pages to the right OCR.**
|
| 37 |
-
extracts case metadata;
|
| 38 |
-
structure; a judicial form to a form-trained OCR.
|
| 39 |
-
this downstream work. It only classifies.
|
| 40 |
2. **Metadata for chunking.** A secondary use is tagging extracted text to help
|
| 41 |
an LLM when working with documents. E.g., use it to tag pages of an exhibit
|
| 42 |
as an exhibit page.
|
|
@@ -50,7 +50,8 @@ Trained on about 13K pages of human-labelled page images. Each page image was
|
|
| 50 |
OCR'd five times, by five different OCR engines. Play around with the tabs
|
| 51 |
below — the same cover page, five very different transcriptions, one label.
|
| 52 |
Note that each page is part of a document, and the whole-document mode below
|
| 53 |
-
takes previous/subsequent pages into account
|
|
|
|
| 54 |
|
| 55 |
<img src="docs/examples/cover_page.jpg" alt="cover page" width="300" align="right">
|
| 56 |
|
|
|
|
| 33 |
|
| 34 |
## Purpose
|
| 35 |
|
| 36 |
+
1. **Routing pages to the right OCR.** Cover page text would go to a model
|
| 37 |
+
that extracts case metadata; TOC-designated to one that reads heading
|
| 38 |
+
structure; text from a judicial form page to a form-trained OCR. (BUT this tool
|
| 39 |
+
doesn't do any of this downstream work. It only classifies.)
|
| 40 |
2. **Metadata for chunking.** A secondary use is tagging extracted text to help
|
| 41 |
an LLM when working with documents. E.g., use it to tag pages of an exhibit
|
| 42 |
as an exhibit page.
|
|
|
|
| 50 |
OCR'd five times, by five different OCR engines. Play around with the tabs
|
| 51 |
below — the same cover page, five very different transcriptions, one label.
|
| 52 |
Note that each page is part of a document, and the whole-document mode below
|
| 53 |
+
takes previous/subsequent pages into account, which mostly matters for edge
|
| 54 |
+
(low-confidence) cases.
|
| 55 |
|
| 56 |
<img src="docs/examples/cover_page.jpg" alt="cover page" width="300" align="right">
|
| 57 |
|