ck37 commited on
Commit
033a712
1 Parent(s): 91357e5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -2
README.md CHANGED
@@ -14,6 +14,7 @@ task_categories:
14
  task_ids:
15
  - hate-speech-detection
16
  - sentiment-classification
 
17
  pretty_name: measuring-hate-speech
18
  tags:
19
  - arxiv:2009.10277
@@ -30,13 +31,13 @@ tags:
30
 
31
  # Dataset card for _Measuring Hate Speech_
32
 
33
- This is a public release of the dataset described in Kennedy et al. (2020), consisting of 39,565 comments annotated by 7,912 annotators, for 135,556 combined rows. The primary outcome variable is the "hate speech score" but the 10 constituent labels (sentiment, (dis)respect, insult, humiliation, inferior status, violence, dehumanization, genocide, attack/defense, hate speech benchmark) can also be treated as outcomes. Includes 8 target identity groups (race/ethnicity, religion, national origin/citizenship, gender, sexual orientation, age, disability, political ideology) and 42 identity subgroups.
34
 
35
  This dataset card is a work in progress and will be improved over time.
36
 
37
  ## Key dataset columns
38
 
39
- * hate_speech_score - continuous hate speech measure, where higher = more hateful and lower = less hateful
40
  * text - lightly processed text of a social media post
41
  * comment\_id - unique ID for each comment
42
  * annotator\_id - unique ID for each annotator
14
  task_ids:
15
  - hate-speech-detection
16
  - sentiment-classification
17
+ - multi-label-classification
18
  pretty_name: measuring-hate-speech
19
  tags:
20
  - arxiv:2009.10277
31
 
32
  # Dataset card for _Measuring Hate Speech_
33
 
34
+ This is a public release of the dataset described in Kennedy et al. (2020) and Sachdeva et al. (2022), consisting of 39,565 comments annotated by 7,912 annotators, for 135,556 combined rows. The primary outcome variable is the "hate speech score" but the 10 constituent ordinal labels (sentiment, (dis)respect, insult, humiliation, inferior status, violence, dehumanization, genocide, attack/defense, hate speech benchmark) can also be treated as outcomes. Includes 8 target identity groups (race/ethnicity, religion, national origin/citizenship, gender, sexual orientation, age, disability, political ideology) and 42 target identity subgroups, as well as 6 annotator demographics and 40 subgroups.
35
 
36
  This dataset card is a work in progress and will be improved over time.
37
 
38
  ## Key dataset columns
39
 
40
+ * hate_speech_score - continuous hate speech measure, where higher = more hateful and lower = less hateful. > 0.5 is approximately hate speech, < -1 is counter or supportive speech, and -1 to +0.5 is neutral or ambiguous.
41
  * text - lightly processed text of a social media post
42
  * comment\_id - unique ID for each comment
43
  * annotator\_id - unique ID for each annotator