steinunnfridriks commited on
Commit
02365f6
·
verified ·
1 Parent(s): d8ad9c9

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +46 -0
README.md ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: openrail
3
+ language:
4
+ - is
5
+ tags:
6
+ - bias-detection
7
+ - icelandic
8
+ - ner
9
+ - socially-responsible-ai
10
+ - prejudice-detection
11
+ - huggingface
12
+ - dataset
13
+ ---
14
+
15
+ # Icelandic Lexicon of Icelandic Bias and Toxicity
16
+
17
+ **Trigger warning:** This lexicon contains biased, offensive, and harmful language.
18
+
19
+ ## Lexicon Description
20
+
21
+ This lexicon is composed of 2,106 Icelandic expressions, including both single words and short multiword phrases, that reflect bias, stigma, or inflammatory language. These expressions were manually compiled from a variety of public sources, including IceTaboo, comment sections on Icelandic blogs and discussion forums, and right-wing Icelandic media platforms. Each expression was categorized in one of 14 bias-related categories: Addiction (101), Disability (95), Origin (288), General (255), LGBTQIA (193), Looks (129), Personal (121), Profanity (86), Religion (209), Sexual (154), Social (175), Stupidity (98), Vulgar (44), and Women (158).
22
+
23
+ ## Intended Uses & Limitations
24
+
25
+ ### Intended Use
26
+ - Research on bias detection in Icelandic text
27
+ - Training and evaluation of bias-aware NLP models
28
+ - Educational purposes for raising awareness of bias in language
29
+
30
+ ### Limitations
31
+ - This lexicon is not an exhaustive list of harmful language in Icelandic.
32
+ - Some words might be appropriate in one context but harmful in another.
33
+
34
+ ## Ethical Considerations
35
+
36
+ This lexicon is released under the **[BigScience OpenRAIL-D License](https://www.licenses.ai/ai-licenses)**, which allows free use with responsible-use restrictions.
37
+ Prohibited uses include:
38
+ - Harassment or discrimination
39
+ - Generating disinformation or hateful content
40
+ - Surveillance targeting individuals or groups
41
+
42
+ The lexicon includes harmful language and should be handled with care. Trigger warnings are recommended for any public deployment.
43
+
44
+ ## Citation
45
+
46
+ Will be added.