dip-bundestag-qa / README.md
wirthual's picture
filtred out referenced answers
33ae0c3
---
task_categories:
- question-answering
language:
- de
tags:
- government
- bundestag
size_categories:
- 10K<n<100K
---
# Dataset Card for Dataset Name
## Dataset Description
- **Homepage:** Quelle: https://dip.bundestag.de/
### Dataset Summary
Extracted the questions and answers from the DIP Service. All PDFs which are used are of type "Antwort".
In the current version the documents are between the following dates:
START_DATE = "2015-05-07"
END_DATE = "2023-07-09"
### Languages
German
## Dataset Structure
Each row of the dataset consists of the following fields: question, answer and document id.
The document id can be used to retrieve the meta data for the underlying PDF file by sending a request to the follwoing endpoint:
https://search.dip.bundestag.de/api/v1/swagger-ui/#/Drucksachen/getDrucksache
### Data Fields
Question
Answer
doc_id
### Data Splits
No split
## Dataset Creation
Download PDF, extract text, Classify entries based on font size, [dehyphenize](https://github.com/pd3f/dehyphen) text, build pairs when possible.
## Dataset Curation
At this point, no complex curation of the dataset was performed.
Answers which simply referred to other answers were filtered out by these regexes:
```
'^Auf die Antwort.*verwiesen.$'
'^Es wird auf die Antwort.*verwiesen.$'
```
#### Who are the source language producers?
https://dip.bundestag.de/
### Licensing Information
Quelle: Deutscher Bundestag/Bundesrat – DIP / "Bundestags-Drucksache"
For further detail see:
https://dip.bundestag.de/documents/nutzungsbedingungen_dip.pdf