Datasets:
ltg
/

Modalities:
Text
Formats:
json
Size:
< 1K
Libraries:
Datasets
pandas
License:
File size: 541 Bytes
51bd338
 
021016f
 
 
 
 
 
 
 
 
51bd338
021016f
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
---
license: cc0-1.0
task_categories:
- conversational
language:
- 'no'
- nb
- nn
pretty_name: NB Samtale  Conversations
size_categories:
- 1K<n<10K
---
# NB Samtale — Conversations

This dataset contains extracted and cleaned conversations from the [NB Samtale corpus](https://www.nb.no/sprakbanken/en/resource-catalogue/oai-nb-no-sbr-85/). The original is a speech corpus made by the Language Bank at the National Library of Norway. The corpus contains orthographically transcribed speech from podcasts and recordings of live events.