Add Blocks and Plots
Browse files- app.py +95 -27
- plot.ipynb +0 -0
app.py
CHANGED
@@ -1,33 +1,85 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
|
3 |
-
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
if methodology == 'A':
|
7 |
-
run_a(dataset_id)
|
8 |
-
elif methodology == 'B':
|
9 |
-
run_b(dataset_id)
|
10 |
-
elif methodology == 'C':
|
11 |
-
run_c(dataset_id)
|
12 |
-
|
13 |
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
with demo:
|
17 |
gr.Markdown("# BiasAware: Dataset Bias Detection")
|
18 |
-
|
|
|
|
|
|
|
19 |
with gr.Row():
|
20 |
with gr.Column(scale=1):
|
21 |
gr.Markdown("Select a dataset to analyze")
|
22 |
|
23 |
-
|
24 |
gr.Examples(
|
25 |
examples=["imdb", "amazon_reviews_multi", "tweet_eval"],
|
26 |
fn=run_evaluation,
|
27 |
-
inputs=[
|
28 |
)
|
29 |
|
30 |
-
methodology = gr.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
|
32 |
button = gr.Button("Run Evaluation")
|
33 |
|
@@ -35,23 +87,39 @@ with demo:
|
|
35 |
gr.Markdown("### Results")
|
36 |
|
37 |
with gr.Box():
|
38 |
-
methodology_title = gr.Markdown("###
|
39 |
methodology_description = gr.Markdown("lorem ipsum")
|
40 |
-
|
41 |
-
|
42 |
-
outputs = gr.Markdown()
|
|
|
|
|
|
|
43 |
gr.Error("No results to display")
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
methodology.change(
|
46 |
-
fn=
|
47 |
inputs=[methodology],
|
48 |
-
outputs=[
|
|
|
|
|
|
|
|
|
49 |
)
|
50 |
|
51 |
-
button.click(
|
52 |
-
|
53 |
-
inputs=[dataset_id, methodology],
|
54 |
-
outputs=[outputs]
|
55 |
-
)
|
56 |
|
57 |
-
demo.launch()
|
|
|
1 |
import gradio as gr
|
2 |
+
import pandas as pd
|
3 |
|
4 |
+
data = [
|
5 |
+
["Category", "Value", "Percentage"],
|
6 |
+
["Total Reviews", 50000, None],
|
7 |
+
["Total Sentences", 621647, None],
|
8 |
+
["Pronouns in Sentences", None, None],
|
9 |
+
["Male Pronouns", 85615, None],
|
10 |
+
["Female Pronouns", 39372, None],
|
11 |
+
["Both Male and Female Pronouns", 7765, None],
|
12 |
+
["Exclusive Usage of Pronouns", None, None],
|
13 |
+
["Only Male Pronouns", 77860, 13.77],
|
14 |
+
["Only Female Pronouns", 31617, 6.33],
|
15 |
+
["Pronouns and Professions in Sentences", None, None],
|
16 |
+
["Male Pronouns with Professions", 5580, 0.9],
|
17 |
+
["Female Pronouns with Professions", 2618, 0.42],
|
18 |
+
["Exclusive Usage of Pronouns with Professions", None, None],
|
19 |
+
["Only Male Pronouns with Professions", 5011, 0.81],
|
20 |
+
["Only Female Pronouns with Professions", 2049, 0.33],
|
21 |
+
["Pronouns and Professions in Combination", None, None],
|
22 |
+
["Male or Female Pronouns with Professions", 7629, 1.23],
|
23 |
+
["Male and Female Pronouns with Professions", 569, 0.09]
|
24 |
+
]
|
25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
+
def display_methodology(methodology):
|
28 |
+
title = methodology
|
29 |
+
description = ""
|
30 |
+
details = ""
|
31 |
+
if methodology == "Term Identity Diversity Analysis":
|
32 |
+
description = "111"
|
33 |
+
details = "222"
|
34 |
+
elif methodology == "Textual Gender Label Evaluation":
|
35 |
+
description = "333"
|
36 |
+
details = "444"
|
37 |
+
elif methodology == "GenBit":
|
38 |
+
description = "555"
|
39 |
+
details = "666"
|
40 |
+
|
41 |
+
return title, description, details
|
42 |
+
|
43 |
+
|
44 |
+
def run_evaluation(dataset, methodology):
|
45 |
+
return f"Running evaluation for {dataset} with {methodology}"
|
46 |
+
|
47 |
+
if methodology == "A":
|
48 |
+
run_a(dataset)
|
49 |
+
elif methodology == "B":
|
50 |
+
run_b(dataset)
|
51 |
+
elif methodology == "C":
|
52 |
+
run_c(dataset)
|
53 |
+
|
54 |
+
|
55 |
+
demo = gr.Blocks(title="BiasAware: Dataset Bias Detection",
|
56 |
+
theme=gr.themes.Soft())
|
57 |
|
58 |
with demo:
|
59 |
gr.Markdown("# BiasAware: Dataset Bias Detection")
|
60 |
+
gr.Markdown(
|
61 |
+
"Natural Language Processing (NLP) training datasets often reflect the biases present in the data sources they are compiled from, leading to the **perpetuation of stereotypes, underrepresentation, and skewed perspectives in AI models**. BiasAware is designed to **identify and quantify biases present in text data**, making it an invaluable resource for data scientists, machine learning practitioners, and organizations committed to **mitigating bias in AI systems**."
|
62 |
+
)
|
63 |
+
|
64 |
with gr.Row():
|
65 |
with gr.Column(scale=1):
|
66 |
gr.Markdown("Select a dataset to analyze")
|
67 |
|
68 |
+
dataset = gr.Text(label="Dataset")
|
69 |
gr.Examples(
|
70 |
examples=["imdb", "amazon_reviews_multi", "tweet_eval"],
|
71 |
fn=run_evaluation,
|
72 |
+
inputs=[dataset],
|
73 |
)
|
74 |
|
75 |
+
methodology = gr.Radio(
|
76 |
+
[
|
77 |
+
"Term Identity Diversity Analysis",
|
78 |
+
"Textual Gender Label Evaluation",
|
79 |
+
"GenBit",
|
80 |
+
],
|
81 |
+
label="Methodology",
|
82 |
+
)
|
83 |
|
84 |
button = gr.Button("Run Evaluation")
|
85 |
|
|
|
87 |
gr.Markdown("### Results")
|
88 |
|
89 |
with gr.Box():
|
90 |
+
methodology_title = gr.Markdown("### Title")
|
91 |
methodology_description = gr.Markdown("lorem ipsum")
|
92 |
+
|
93 |
+
methodology_details = gr.Markdown("lorem ipsum")
|
94 |
+
# outputs = gr.Markdown()
|
95 |
+
outputs = gr.DataFrame(pd.DataFrame(data), headers=[
|
96 |
+
"", "Count", "Percentage"])
|
97 |
+
|
98 |
gr.Error("No results to display")
|
99 |
+
|
100 |
+
with gr.Column(scale=1):
|
101 |
+
gr.Markdown("### Leaderboard")
|
102 |
+
gr.DataFrame(
|
103 |
+
headers=["Dataset", "Score"],
|
104 |
+
value=[
|
105 |
+
["imdb", 0.9],
|
106 |
+
["amazon_reviews_multi", 0.8],
|
107 |
+
["tweet_eval", 0.7],
|
108 |
+
],
|
109 |
+
interactive=False,
|
110 |
+
)
|
111 |
+
|
112 |
methodology.change(
|
113 |
+
fn=display_methodology,
|
114 |
inputs=[methodology],
|
115 |
+
outputs=[
|
116 |
+
methodology_title,
|
117 |
+
methodology_description,
|
118 |
+
methodology_details,
|
119 |
+
],
|
120 |
)
|
121 |
|
122 |
+
button.click(fn=run_evaluation, inputs=[
|
123 |
+
dataset, methodology], outputs=[outputs])
|
|
|
|
|
|
|
124 |
|
125 |
+
demo.launch()
|
plot.ipynb
ADDED
The diff for this file is too large to render.
See raw diff
|
|