File size: 3,755 Bytes
3cc2acc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6368811
3cc2acc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---

annotations_creators:
  - expert-generated
language_creators:
  - expert-generated
language:
  - nb
  - nn
license: cc-by-4.0
multilinguality:
  - translation
pretty_name: Nynorsk Norm 200eval
size_categories:
  - 1K<n<10K
source_datasets:
  - original
task_categories:
  - translation
task_ids:
  - text2text-generation
---


# Nynorsk Norm 200eval

`nynorsk_norm_200eval` is a high-quality, small-scale parallel corpus comprising 200 Norwegian Bokmål–Nynorsk sentence pairs collected from official sources and public institutions. Each example includes:

- `nb`: Original sentence in Bokmål  
- `nn_original`: Original Nynorsk sentence (typically an official translation)  
- `nn_alt_original`: Original Nynorsk sentence (typically an official translation) - alt version  
- `nn_husnorm`: Sentence rewritten in Nynorsk following an internal *husnorm* (house style)  
- `nb_source`, `nn_source`: Source URLs for the Bokmål and Nynorsk versions

## Benchmark

### Target: `nn_original` (official Nynorsk)



| Model            | BLEU | chrF | Notes               |

|------------------|------|------|---------------------|

| ExampleModel-v1  | XX.X | XX.X | Official translation baseline |

| ExampleModel-v2  | XX.X | XX.X | Fine-tuned on government data |



### Target: `nn_husnorm` (house style Nynorsk)

| Model            | BLEU | chrF | Notes                     |
|------------------|------|------|---------------------------|
| ExampleModel-v1  | XX.X | XX.X | Without style normalization |
| ExampleModel-v2  | XX.X | XX.X | Fine-tuned on husnorm corpus |

> Will be replaced with real numbers as soon as we have performed the tests.


## Intended Uses

This dataset is intended for:

- Evaluation of Bokmål-to-Nynorsk translation systems  
- Research on Nynorsk stylistic variation  
- Style transfer and normalization experiments, especially in public communication

## Format

The dataset is provided as a UTF-8 encoded JSON Lines (`.jsonl`) file. Each line is a structured example:
```json

{

  "nb": "Ledelsen ved NMBU disponerer i dag to elbiler som de kan bruke på møter i nærområdet.",

  "nn_original": "Leiinga ved NMBU disponerer i dag to elbilar som dei kan bruka på møte i nærområdet.",

  "nn_husnorm": "Leiinga ved NMBU disponerer i dag to elbilar som dei kan bruke på møte i nærområdet.",

  "nb_source": "https://www.nmbu.no/om/miljoarbeidet/milj-rsrapport-2020",

  "nn_source": "https://www.nmbu.no/om/miljoarbeidet/miljotiltak"

}

```

## Nynorsk house style ("husnorm")

The `nn_husnorm` field contains Nynorsk text rewritten according to an internal style guide.
This house norm favors Bokmål-near variants for improved consistency and accessibility in public texts.

General characteristics:
- Use of samsvarsbøying (agreement)
- Prefer active voice where natural
- Rewrite s-genitive using prepositions, compounds, or possessives (sin/si)
- Use e-infinitive: å komme (not å kome)
- Use bli, blei, blitt (not verte, vart, vorte)
- Pronouns: vi, ho, henne (not me, ho, henne)
- Adverbs: nå, da, derfor, fordi (not no, då, difor, av di)
- Noun plural forms:
  - gjestar, nadar instead of gjester, nader
  - elver, helger instead of elvar, helgar
- Bokmål-like variants of verbs and nouns:
  - følge, givar, følgar over følgje, gjevar, følgjar
- Prefer forelesing, foreslå over førelesing, føreslå
- Use doubled consonants: komme, lønne (not kome, løne)
- Lexical choices like nødvendig, allmente are accepted

## Authors

- Husnorm adaptation and linguistic quality control: **Arne Martinus Lidstad** and **Marie Røsok**  
- Dataset formatting and Hugging Face integration: **Per Egil Kummervold**