TeetouchQQ
commited on
Commit
•
d0f56a6
1
Parent(s):
423eeb1
Update README.md
Browse files
README.md
CHANGED
@@ -11,34 +11,86 @@ Typhoon Safety is a lightweight binary classifier designed to detect harmful con
|
|
11 |
|
12 |
Train on mixed of Thai Sensitive topic dataset and Wildguard.
|
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 |
## Model Details
|
44 |
|
@@ -68,12 +120,10 @@ Train on mixed of Thai Sensitive topic dataset and Wildguard.
|
|
68 |
|
69 |
|
70 |
- **Developed by:** [More Information Needed]
|
71 |
-
- **
|
72 |
-
- **
|
73 |
-
- **
|
74 |
-
- **
|
75 |
-
- **License:** [More Information Needed]
|
76 |
-
- **Finetuned from model [optional]:** [More Information Needed]
|
77 |
|
78 |
|
79 |
## How to Get Started with the Model
|
|
|
11 |
|
12 |
Train on mixed of Thai Sensitive topic dataset and Wildguard.
|
13 |
|
14 |
+
this model is trained to predict safety labels on below categories.
|
15 |
+
|
16 |
+
<div class="section-header">Thai Sensitive Topics</div>
|
17 |
+
<table align="center">
|
18 |
+
<tr>
|
19 |
+
<th colspan="3">Category</th>
|
20 |
+
</tr>
|
21 |
+
<tr>
|
22 |
+
<td>The Monarchy</td>
|
23 |
+
<td>Student Protests and Activism</td>
|
24 |
+
<td>Drug Policies</td>
|
25 |
+
</tr>
|
26 |
+
<tr>
|
27 |
+
<td>Gambling</td>
|
28 |
+
<td>Cultural Appropriation</td>
|
29 |
+
<td>Thai-Burmese Border Issues</td>
|
30 |
+
</tr>
|
31 |
+
<tr>
|
32 |
+
<td>Cannabis</td>
|
33 |
+
<td>Human Trafficking</td>
|
34 |
+
<td>Military and Coup/td>
|
35 |
+
</tr>
|
36 |
+
<tr>
|
37 |
+
<td>LGBTQ+ Rights</td>
|
38 |
+
<td>Political Divide</td>
|
39 |
+
<td>Religion and Buddhism</td>
|
40 |
+
</tr>
|
41 |
+
<tr>
|
42 |
+
<td>Political Corruption</td>
|
43 |
+
<td>Foreign Influence</td>
|
44 |
+
<td>National Identity and Immigration</td>
|
45 |
+
</tr>
|
46 |
+
<tr>
|
47 |
+
<td>Freedom of Speech and Censorship</td>
|
48 |
+
<td>Vape</td>
|
49 |
+
<td>Southern Thailand Insurgency</td>
|
50 |
+
</tr>
|
51 |
+
<tr>
|
52 |
+
<td>Sex Tourism and Prostitution</td>
|
53 |
+
<td>COVID-19 Management</td>
|
54 |
+
<td>Royal Projects and Policies</td>
|
55 |
+
</tr>
|
56 |
+
<tr>
|
57 |
+
<td>Migrant Labor Issues</td>
|
58 |
+
<td>Environmental Issues and Land Rights</td>
|
59 |
+
<td></td>
|
60 |
+
</tr>
|
61 |
+
</table>
|
62 |
+
|
63 |
+
<div class="section-header">Wildguard Topics</div>
|
64 |
+
<table>
|
65 |
+
<tr>
|
66 |
+
<th colspan="3">Category</th>
|
67 |
+
</tr>
|
68 |
+
<tr>
|
69 |
+
<td>Others</td>
|
70 |
+
<td>Sensitive Information Organization</td>
|
71 |
+
<td>Mental Health Over-reliance Crisis</td>
|
72 |
+
</tr>
|
73 |
+
<tr>
|
74 |
+
<td>Social Stereotypes & Discrimination</td>
|
75 |
+
<td>Defamation & Unethical Actions</td>
|
76 |
+
<td>Cyberattack</td>
|
77 |
+
</tr>
|
78 |
+
<tr>
|
79 |
+
<td>Disseminating False Information</td>
|
80 |
+
<td>Private Information Individual</td>
|
81 |
+
<td>Copyright Violations</td>
|
82 |
+
</tr>
|
83 |
+
<tr>
|
84 |
+
<td>Toxic Language & Hate Speech</td>
|
85 |
+
<td>Fraud Assisting Illegal Activities</td>
|
86 |
+
<td>Causing Material Harm by Misinformation</td>
|
87 |
+
</tr>
|
88 |
+
<tr>
|
89 |
+
<td>Violence and Physical Harm</td>
|
90 |
+
<td>Sexual Content</td>
|
91 |
+
<td></td>
|
92 |
+
</tr>
|
93 |
+
</table>
|
94 |
|
95 |
## Model Details
|
96 |
|
|
|
120 |
|
121 |
|
122 |
- **Developed by:** [More Information Needed]
|
123 |
+
- **Model type:** Transformer Encoder
|
124 |
+
- **Language(s) (NLP):** Thai 🇹🇠and English 🇬🇧
|
125 |
+
- **License:** MIT
|
126 |
+
- **Finetuned from model [optional]:** mDeBERTa v3 base https://huggingface.co/microsoft/mdeberta-v3-base
|
|
|
|
|
127 |
|
128 |
|
129 |
## How to Get Started with the Model
|