README.md

---
license: mit
language:
- de
tags:
- QA
dataset_info:
  features:
  - name: id
    dtype: string
  - name: title
    dtype: string
  - name: context
    dtype: string
  - name: question
    dtype: string
  - name: answers
    struct:
    - name: answer_start
      sequence: int64
    - name: text
      sequence: string
  splits:
  - name: test
    num_bytes: 560403
    num_examples: 357
  - name: train
    num_bytes: 2826731
    num_examples: 1773
  download_size: 710027
  dataset_size: 3387134
task_categories:
- question-answering
pretty_name: 'LHM Dienstleistungen: QA'
size_categories:
- 1K<n<10K
---
# LHM-Dienstleistungen-QA - german public domain question-answering dataset 
Datasets created based on data from Munich city administration.
Format inspired by GermanQuAD. 
  
## Annotated by: 
-       Institute for Applied Artificial Intelligence: Leon Marius Schröder 
-       BettercallPaul GmbH: Clemens Gutknecht, Oubada Alkiddeh, Susanne Weiß 
-       Stadt München: Leon Lukas 
  
  
  
## Data basis 
Texts taken from the “Dienstleistungsfinder“ of the city of Munich administration. 
There information about services offered by city is presented online.
Information ranges from applying for an ID card to dispose of garbage.
- https://stadt.muenchen.de/service/ (Date 11/2022)
  
## Dataset statistics 
-       Shortest Question: 13 Characters 
-       Average Question: 68 Characters 
-       Longest Question: 183 Characters 
  
### Distribution of first sentence beginnings 
![all_words](alle.jpg " All sentence beginnings ") 
### Distribution of first sentence beginnings: Wie 
![Wie](Wie.jpg " Wie sentence beginnings") 
### Distribution of first sentence beginnings: Wo 
![Wo](wo.jpg " Wo sentence beginnings") 
### Distribution of first sentence beginnings: Was 
![Was](Was.jpg " Was sentence beginnings")

## Models trained using this datset
### QA
- cgutknecht/gelectra_large_gsqd-gq-LHM
### DPR 
- schreon/xnext-lhm_queries_encoder
- schreon/xnext-lhm_passages_encoder