README.md

---
license: mit
dataset_info:
  features:
  - name: name
    dtype: string
  - name: language
    dtype: string
  - name: description
    dtype: string
  - name: source
    dtype: string
  - name: public
    dtype: bool
  - name: id
    dtype: int64
  - name: lastModification
    dtype: timestamp[us, tz=UTC]
  - name: keywords
    sequence: string
  - name: embedding
    sequence: float64
  - name: content
    dtype: string
  splits:
  - name: train
    num_bytes: 43495906
    num_examples: 1446
  download_size: 39682949
  dataset_size: 43495906
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
task_categories:
- feature-extraction
language:
- de
- en
tags:
- government
- german
- munich
- services
- administration
- citizen
pretty_name: Munich Public Services / Münchner Verwaltungs-Dienstleistungen
size_categories:
- 1K<n<10K
---
# Dataset Card for "Munich Public Services"

[**zur deutschen Dokumentation**](https://ki.muenchen.de/datensaetze/munich-public-services.html)

<!-- Provide a quick summary of the dataset. -->
This dataset contains information about the services provided to the public by the City of Munich in the form of written articles, corresponding metadata as well as embeddings.

## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->
The Munich Public Services dataset contains around 1.400 articles about various public services provided by the City of Munich.
Next to the content of the articles, the dataset also includes metadata such the language of the article, the last modification date, and the keywords associated with the article.
The dataset also includes embedding vectors computed from each article, which can be used for Retrieval or other NLP tasks.

- **Curated by:** City of Munich, it@M, KI Competence Center
- **Language(s):** German, English
- **License:** MIT

### Dataset Sources

<!-- Provide the basic links for the dataset. -->
- **City of Munich AI Documentation:** [ki.muenchen.de](https://ki.muenchen.de/)
- **Dataset Source Webpages:** [Munich Public Services](https://stadt.muenchen.de/service/en-GB/)

## Uses

<!-- Address questions around how the dataset is intended to be used. -->

### Direct Use

<!-- This section describes suitable use cases for the dataset. -->
The dataset can be used to provide information about the public services provided by the City of Munich.
This is especially useful for grounding information retrieval systems or Large Language Models in the context of public services.
The data can also be used to analyze itself for patterns or trends in the public services provided by the City of Munich.

### Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
The dataset should not be used for any purpose other than the intended use case of providing information about the public services provided by the City of Munich.
This also includes other cities or government entities, as the information is specific to Munich.

The dataset should not be used for any malicious purposes like desinformation or fake news generation.
Also using the dataset as input to an LLM for generation of more 'fake' articles should not be done.

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
The dataset consists of the following fields:

| Field             | Type       | Description                                                                                   |
|-------------------|------------|-----------------------------------------------------------------------------------------------|
| `id`              | int      | The unique identifier of the article.                                                         |
| `name`            | string     | The title of the article.                                                                     |
| `content`         | string     | The full text content of the article, formatted in Markdown.                                  |
| `language`        | string     | The language of the article, either German (`de`) or English (`en`).                          |
| `description`     | string     | A short description of the article content with a maximum length of x characters, formatted in HTML. |
| `source`          | string     | Link to the source webpage of the article.                                                    |
| `public`          | bool       | Boolean value indicating if the article is public or not, always `True`.                      |
| `lastModification`| timestamp  | Timestamp of the last modification of the article in UTC.                                     |
| `keywords`        | sequence [string]   | A list of keywords associated with the article, the keywords might be in German or English.   |
| `embedding`       | sequence [float]  | A 3072 dimensional embedding vector computed from the article content, using OpenAI's [`text-embedding-3-large`](https://platform.openai.com/docs/guides/embeddings/#embedding-models) model. |

## Dataset Creation

### Curation Rationale

<!-- Motivation for the creation of this dataset. -->
The dataset was created as a side product of a project to provide a better information retrieval system for the City of Munich's public services.
Periodically the dataset is updated to include new articles and to update the embeddings.

### Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
The dataset is sourced from the city of Munich's public services website and it's underlying CMS.
All articles were written by the city of Munich's employees.

#### Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
The data was collected by querying an internal API of the city of Munich's CMS.
The data was then processed to extract the metadata and the content of the articles.
The article content was transformed from HTML to Markdown by using the [markdownify](https://pypi.org/project/markdownify/) library.
The embeddings were generated using OpenAI's [`text-embedding-3-large`](https://platform.openai.com/docs/guides/embeddings/#embedding-models) model.

#### Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
The source data producers are the employees of the City of Munich who wrote the articles, there is no attribution to individual authors.
The employees are a diverse group of people working in the city's administration with special knowledge about the public services of their respective departments and skills in writing understandable articles.

#### Personal and Sensitive Information

<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
The dataset does not contain any personal or sensitive information.

## Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->
The provided dataset is only a subset of all public services provided by the City of Munich.
Besides that, the data can contain biases in the form of

- the language used in the articles (so called "Beamtendeutsch" / "administrative German")
- the quality and information richness of the articles
- the number of articles for one topic
- and the keywords associated with the articles.

This can lead to biases in the embeddings and the information retrieval system built on top of the dataset.

### Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
To mitigate the biases in the dataset, it is recommended to use the dataset in combination with other sources of information.
Another approach is to evaluate the dataset with a custom evaluation set to identify biases for the specific use case and to mitigate them.

## More Information

For more information about the AI projects of the City of Munich, please visit the [ki.muenchen.de](https://ki.muenchen.de/) website.

## Dataset Card Authors

Fabian Reinold, City of Munich, it@M, AI Competence Center

## Dataset Card Contact

For any questions about the dataset, please [contact us](mailto:[email protected])