File size: 4,762 Bytes
ed22468
 
 
43d839a
ed22468
 
 
43d839a
 
ed22468
43d839a
65faa6e
2debafd
43d839a
 
ed22468
 
060c443
ed22468
 
65faa6e
ed22468
 
 
 
 
 
 
 
 
 
 
 
2c91001
 
ed22468
 
 
 
 
d36009a
2c91001
6a2c8fb
 
ed22468
 
 
 
 
 
 
 
 
 
2c91001
6a2c8fb
ed22468
2c91001
 
 
ed22468
 
 
 
 
 
 
43d839a
ed22468
 
 
 
43d839a
ed22468
 
 
 
 
 
43d839a
ed22468
43d839a
ed22468
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
43d839a
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
---
pretty_name: Danske Taler
language:
- da
license: cc0-1.0
license_name: CC-0
task_categories:
- text-generation
- fill-mask
task_ids:
- language-modeling
domains:
- Conversation
- Speeches
- Spoken
---

# Dataset Card for Danske Taler

<!-- START-SHORT DESCRIPTION -->
Danish Speeches from [dansketaler.dk](https://www.dansketaler.dk).
<!-- END-SHORT DESCRIPTION -->


The database dansketaler.dk is managed by Danske Taler, an independent institution that in addition to managing the database and carries out cultural
and democratic projects based on speeches. 
Danske Taler state as their goals that they seek to preserve our cultural heritage and promotes active citizenship and democratic confidence through its work. 
Additionally, Danske Taler provides data to a number of online resources, including: lex.dk, sprogteknologi.dk, and ordnet.dk.

The goal of the dataset is to collect historical and timely speeches and make them available for the public.

Learn more about danske taler by reading their [about us](https://www.dansketaler.dk/om-os) page.

> NOTE: Danske-Taler is also collecting [sermons](https://www.dansketaler.dk/praedikener), but these are not included in this dataset. 

## Dataset Description


<!-- START-DESC-STATS -->
- **Language**: dan, dansk, Danish
- **Domains**: Conversation, Speeches, Spoken
- **Number of samples**: 2.91K
- **Number of tokens (Llama 3)**: 8.72M
- **Average document length (characters)**: 9140.42
<!-- END-DESC-STATS -->


## Dataset Structure
An example from the dataset looks as follows.


<!-- START-SAMPLE -->
```py
{
  "id": "danske-taler_281",
  "text": "Tyske landsmænd og -kvinder !\nSyv år er kort tid, en brøkdel af en enkel menneskelig normaltilværels[...]",
  "source": "danske-taler",
  "added": "2025-06-24",
  "created": "1940-01-30, 1940-01-30",
  "token_count": 3020
}
```

### Data Fields

An entry in the dataset consists of the following fields:

- `id` (`str`): An unique identifier for each document.
- `text`(`str`): The content of the document.
- `source` (`str`): The source of the document (see [Source Data](#source-data)).
- `added` (`str`): An date for when the document was added to this collection.
- `created` (`str`): An date range for when the document was originally created.
- `token_count` (`int`): The number of tokens in the sample computed using the Llama 8B tokenizer
<!-- END-SAMPLE -->


### Dataset Statistics

<!-- START-DATASET PLOTS -->
<p align="center">
<img src="./images/dist_document_length.png" width="600" style="margin-right: 10px;" />
</p>
<!-- END-DATASET PLOTS -->



## Additional Information


### Dataset Collection Process

This dataset was collected using the publicly available [API](https://www.dansketaler.dk/api/v1). 

### Quality Assurance
We check for and remove exact duplicates, empty texts, duplicate ids after the initial download. We additionally check if the articles contain any HTML.

## Opportunities for Improvement

While this dataset can be updated to include the latest availabe speeches. 

We consider the quality of the current collection high with a low chance of 
incorrect formatting, 
spelling errors,
empty documents or 
misformatted segments. 
This stems both from the quality assurance, source of documents and subjective inspection.

### License Information
Since the license information isn't avaiable through the API we collect this data directly from the webpage of each article under the header 
"Ophavsret".

For speeches where it is noted that *"Materialet er fri af ophavsret"* (The material is in the public domain) or similarly we assign it a `cc0` license.

Such an example can be seen here:

> **Ophavsret**
> 
> Materialet er fri af ophavsret. Taler, som er holdt i offentligheden, er ikke omfattet af ophavsret (Jf. ophavsretslovens § 26 og 32). 
> Det betyder, at når en tale er indgået i Danske Talers database, kan den bruges af tredjeparter, fx til undervisning eller forskning.
>
> *source: [Ursula von der Leyens tale om europæisk forsvar og sikkerhed på Hærens Officersskole](https://www.dansketaler.dk/tale/tale-om-europaeisk-forsvar-og-sikkerhed-pa-haerens-officersskole)*

Speeches without this mention is removed. Such an example include:

> **Ophavsret**
> 
> Materialet er beskyttet af ophavsret
>
> *Source: [Christina Egelunds tale ved Aarhus Universitets årsfest](https://www.dansketaler.dk/tale/christina-egelunds-tale-ved-aarhus-universitets-arsfest)*

We manually checked the unique set of license descriptions to see if any were open licenses that weren't included in the current criteria.

For specific filtering criteria see the `create.py` script.

### Citation Information

No citation is applicable for this work. We recommend citing the huggingface repository.