File size: 2,775 Bytes
79c1f88
 
 
 
 
 
 
 
 
 
5cc15c7
79c1f88
 
 
 
 
 
a9ae332
79c1f88
c9930b1
 
79c1f88
5cc15c7
c9930b1
79c1f88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9e4bb37
79c1f88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5cc15c7
 
2748e60
5cc15c7
 
 
c05d6c8
2748e60
 
 
 
5cc15c7
 
 
 
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
---
license: apache-2.0
task_categories:
- text-classification
language:
- pt
- en
- de
- es
- tr
- fr
pretty_name: MOL
---
# MOL - Context-Aware Multilingual Offensive Lexicon

The MOL is the first specialized lexicon for hate speech detection, annotated with contextual information.

<b>It consists of 1,000 explicit and implicit (clue-based) human-annotated rationales</b> used with pejorative connotations, manually identified by a linguist and annotated by three experts regarding their contextual dependency (context-dependent or context-independent).

For example, the term "stupid" is classified as a <b>context-independent offensive term</b>, as it is predominantly used in pejorative contexts.
In contrast, the terms "useless" and "worm" are considered <b>context-dependent offensive terms</b> because they can appear in both non-pejorative contexts—such as "this smartphone is useless" or "the fisherman uses worms for bait"—and pejorative contexts, such as "this last President was useless" or "this human being is such a worm."

Each term and expression in the MOL was annotated by three independent annotators, achieving a high inter-annotator agreement score (73% Kappa).
Originally developed in Portuguese, MOL was manually translated by native speakers into <b>English, Spanish, French, German, and Turkish</b>, ensuring cultural adaptations in each language. As a result, MOL is available in six different languages.



The table below describes the MOL statistics
<div align="center">
<table> 
<tr><th>Contextual Information</th><th>Hate Targets </th></tr>
<tr><td>

|class|label|total|
|--|--|--|  
|Context-independent offensive|1|612| 
|Context-depedent offensive|0|387| 
 |Total||1,000| 


</td><td>

|class|total|  
|--|--|  
|non-target |864|
|partyism|69|
|sexism|35|
|homophobia|16|
|fatphobia|9|
|religious intolerance|9|
|antisemitism|1|
|apology for the dictatorship|5|
|racism|4|  
|antisemitistm|3| 
|total|1,000|


</td></tr></table>
</div>


# Dataset Description
- **Created by:** Francielle Vargas (<https://franciellevargas.github.io/>)
- **Funded by:** FAPESP 
- **Language(s) (NLP):** Portuguese, English, Spanish, French, Germany and Turkish
- **Repository:** https://github.com/franciellevargas/MOL
- **Paper:** <br>
<b>Context-Aware and Expert Data Resources for Brazilian Portuguese Hate Speech Detection</b> <br>
Francielle Vargas, Isabelle Carvalho, Thiago A.S. Pardo, Fabrício Benevenuto</i> <br>
<i>Natural Language Engineering Journal</i>. Cambridge University Press. 2024. <br>
https://www.cambridge.org/core/journals/natural-language-processing/article/contextaware-and-expert-data-resources-for-brazilian-portuguese-hate-speech-detection/7D9019ED5471CD16E320EBED06A6E923#


# Dataset Contact
francielealvargas@gmail.com