nielsr HF Staff commited on
Commit
2afa34f
·
verified ·
1 Parent(s): 06ab9e6

Improve dataset card: Add task categories, tags, paper link, and GitHub link

Browse files

This PR enhances the dataset card by:
- Adding `task_categories: ['object-detection']` to the metadata, as form field detection is an object detection task.
- Adding relevant `tags` such as `document-intelligence`, `form-field-detection`, `pdf`, and `multilingual` for better discoverability and context.
- Providing a link to the paper ([CommonForms: A Large, Diverse Dataset for Form Field Detection](https://huggingface.co/papers/2509.16506)) for easy access to the research.
- Including a link to the GitHub repository ([https://github.com/jbarrow/commonforms](https://github.com/jbarrow/commonforms)) for the associated code.
- Adding a comprehensive description of the dataset based on the paper abstract.

Files changed (1) hide show
  1. README.md +25 -3
README.md CHANGED
@@ -1,3 +1,25 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - object-detection
5
+ tags:
6
+ - document-intelligence
7
+ - form-field-detection
8
+ - pdf
9
+ - multilingual
10
+ ---
11
+
12
+ # CommonForms: A Large, Diverse Dataset for Form Field Detection
13
+
14
+ This repository hosts the **CommonForms** dataset, a web-scale dataset for form field detection, introduced in the paper [CommonForms: A Large, Diverse Dataset for Form Field Detection](https://huggingface.co/papers/2509.16506).
15
+
16
+ CommonForms casts the problem of form field detection as object detection: given an image of a page, predict the location and type (Text Input, Choice Button, Signature) of form fields.
17
+
18
+ **Key Features:**
19
+ * **Scale:** Roughly 55,000 documents comprising over 450,000 pages.
20
+ * **Source:** Constructed by filtering Common Crawl to find PDFs with fillable elements.
21
+ * **Diversity:** Contains a diverse mixture of languages (one third non-English) and domains, with no single domain making up more than 25% of the dataset.
22
+ * **Purpose:** The first large-scale dataset released for form field detection, aimed at fostering the development of robust form field detectors.
23
+
24
+ **Paper:** [CommonForms: A Large, Diverse Dataset for Form Field Detection](https://huggingface.co/papers/2509.16506)
25
+ **Code:** [https://github.com/jbarrow/commonforms](https://github.com/jbarrow/commonforms)