|
--- |
|
license: mit |
|
language: |
|
- cs |
|
tags: |
|
- document question answering |
|
--- |
|
|
|
# LayoutXLM Model Fine-tuned with CIVQA and DVQA dataset |
|
|
|
This is a fine-tuned version of the [LayoutXLM model](https://huggingface.co/microsoft/layoutxlm-base), which was trained on Czech Invoice Visual Question Answering (CIVQA) dataset containing invoices in the Czech language as well as on the Data Visualizations via Question Answering ([DVQA] (https://paperswithcode.com/dataset/dvqa)) dataset. |
|
|
|
This model enables Document Visual Question Answering on Czech invoices with the use of the existing DVQA dataset. |
|
|
|
Regarding the Czech invoices, we focused on 10 different entities, which are crucial for processing the invoices. |
|
- Variable symbol |
|
- Specific symbol |
|
- Constant symbol |
|
- Bank code |
|
- Account number |
|
- Total amount |
|
- Invoice date |
|
- Name of supplier |
|
- DIC |
|
- QR code |
|
|
|
You can find more information about this model in this [paper](https://nlp.fi.muni.cz/raslan/raslan23.pdf#page=31). |
