Commit
·
d67ae66
1
Parent(s):
e1417b6
Update README.md
Browse files
README.md
CHANGED
|
@@ -67,39 +67,55 @@ The perturber is a finetuned BART model (Lewis et al., 2020) with 24 layers, 102
|
|
| 67 |
|
| 68 |
This model release is separately trained using the HuggingFace transformers library, with the same parameters as the ParlAI model.
|
| 69 |
|
| 70 |
-
Currently the perturber supports text rewriting along three axes and several attributes:
|
| 71 |
-
- **gender:** man, woman, non-binary
|
| 72 |
-
- **race:** black, white, asian, hispanic, native-american, pacific-islander
|
| 73 |
-
- **age:** child, young, middle-aged, senior, adult
|
| 74 |
-
|
| 75 |
### Uses
|
| 76 |
|
| 77 |
The perturber is intended for use by fairness researchers and engineers working on demographic debiasing applications. The perturber is a controllable generation model that given a word, target demographic attribute and input text, outputs text where the selected word and associated references are rewritten to the target demographic attribute. Control variables and the input text are separated by a <PERT_SEP> token.
|
| 78 |
|
| 79 |
## Examples
|
|
|
|
|
|
|
|
|
|
| 80 |
|
| 81 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
his, woman <PERT_SEP> Jack was passionate about rock climbing and his love for the sport was infectious to all men around him.
|
| 83 |
|
| 84 |
-
|
| 85 |
-
Jackie was passionate about rock climbing and her love for the sport was infectious to all men around her.
|
|
|
|
| 86 |
|
| 87 |
-
|
| 88 |
Alice, man <PERT_SEP> To her girlfriend Jen, Alice was a doting mother, loving girlfriend and talented actress.
|
| 89 |
|
| 90 |
-
|
| 91 |
To his girlfriend Jen, Alan was a doting father, loving partner and talented actor.
|
| 92 |
-
|
| 93 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
child, senior <PERT_SEP> The young child is naive and his innocence must be protected at all costs.
|
| 95 |
|
| 96 |
-
|
| 97 |
The elderly person is naive and his innocence must be protected at all costs.
|
| 98 |
|
| 99 |
-
|
|
|
|
| 100 |
Asian, black <PERT_SEP> The Asian students association often hosted anime nights and boba events on campus.
|
| 101 |
|
| 102 |
-
|
| 103 |
The Black students association often hosted anime nights and boba events on campus.
|
| 104 |
|
| 105 |
### Contributions
|
|
|
|
| 67 |
|
| 68 |
This model release is separately trained using the HuggingFace transformers library, with the same parameters as the ParlAI model.
|
| 69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
### Uses
|
| 71 |
|
| 72 |
The perturber is intended for use by fairness researchers and engineers working on demographic debiasing applications. The perturber is a controllable generation model that given a word, target demographic attribute and input text, outputs text where the selected word and associated references are rewritten to the target demographic attribute. Control variables and the input text are separated by a <PERT_SEP> token.
|
| 73 |
|
| 74 |
## Examples
|
| 75 |
+
Below we show some example inputs and outputs for the perturber rewriting text along different demographic axes and attributes.
|
| 76 |
+
|
| 77 |
+
Model inputs follow the format `[selected_word][target_attribute] <PERT_SEP> [input_text]`, where `selected_word` is a word that contains demographic information, `target_attribute` is a demographic attribute such as "man" or "asian", and `input_text` is the text sequence to rewrite.
|
| 78 |
|
| 79 |
+
Currently the perturber supports text rewriting along three axes and several attributes:
|
| 80 |
+
- **gender:** man, woman, non-binary
|
| 81 |
+
- **race:** black, white, asian, hispanic, native-american, pacific-islander
|
| 82 |
+
- **age:** child, young, middle-aged, senior, adult
|
| 83 |
+
|
| 84 |
+
### Gender
|
| 85 |
+
_Input:_
|
| 86 |
his, woman <PERT_SEP> Jack was passionate about rock climbing and his love for the sport was infectious to all men around him.
|
| 87 |
|
| 88 |
+
_Output:_
|
| 89 |
+
Jackie was passionate about rock climbing and her love for the sport was infectious to all men around her.
|
| 90 |
+
<br/>
|
| 91 |
|
| 92 |
+
_Input:_
|
| 93 |
Alice, man <PERT_SEP> To her girlfriend Jen, Alice was a doting mother, loving girlfriend and talented actress.
|
| 94 |
|
| 95 |
+
_Output:_
|
| 96 |
To his girlfriend Jen, Alan was a doting father, loving partner and talented actor.
|
| 97 |
+
<br/>
|
| 98 |
+
<br/>
|
| 99 |
+
|
| 100 |
+
_Input:_
|
| 101 |
+
his, non-binary <PERT_SEP> Jack was passionate about rock climbing and his love for the sport was infectious to all men around him.
|
| 102 |
+
|
| 103 |
+
_Output:_
|
| 104 |
+
Jack was passionate about rock climbing and their love for the sport was infectious to all men around them.
|
| 105 |
+
<br/>
|
| 106 |
+
<br/>
|
| 107 |
+
### Age
|
| 108 |
+
_Input:_
|
| 109 |
child, senior <PERT_SEP> The young child is naive and his innocence must be protected at all costs.
|
| 110 |
|
| 111 |
+
_Output:_
|
| 112 |
The elderly person is naive and his innocence must be protected at all costs.
|
| 113 |
|
| 114 |
+
### Race/Ethnicity
|
| 115 |
+
_Input:_
|
| 116 |
Asian, black <PERT_SEP> The Asian students association often hosted anime nights and boba events on campus.
|
| 117 |
|
| 118 |
+
_Output:_
|
| 119 |
The Black students association often hosted anime nights and boba events on campus.
|
| 120 |
|
| 121 |
### Contributions
|