Is the annotation rule same as the FUNSD?

[tengerye]

Hi, may I ask if the annotation rule is the same as FUNSD? There are a number of strange examples.

[Isydmr]

Thank you for sharing the dataset @ranpox @wolfshow.

I agree with @tengerye.

I'm trying to annotate my custom dataset to train with XFUND.

I made visualizations on the existing datasets (XFUND and FUNDS) to properly understand annotation rules. In some cases, I encountered confusing labeling. Here are some remarks:

_{Orange, blue, green, and pink boxes correspond to the header, question, answer, and other respectively. Linking array displayed on top of boxes.}

a) Question or Header?

FUNSD

XFUND

Headers do not contain linking arrays in XFUND whereas FUNSD contains a lot of header with linking array that points to questions.

---

b) Unanswered Questions

FUNSD

XFUND

XFUND docs labeled such as unanswered questions tagged as Other (see 'Téléphone 2')

c) Labeling Inconsistencies

XFUND - 1

XFUND - 2

IDENTITE is tagged as question but Identité du candidat is tagged as other in different images from XFUND.

[wolfshow]

@Isydmr We are working on XFUND v1.1. Please stay tuned.

[Isydmr]

What is the timeline for XFUND v1.1? We have a tagging project and we were planning to rely on XFUND v1.0 document annotation rules. Should we wait for XFUND v1.1?

[hasansalimkanmaz]

+1

[Rijgersberg]

Any idea on the timeline for XFUND v1.1? I'm considering annotating a Dutch dataset in XFUND style.