Update app.py
Browse files
app.py
CHANGED
@@ -35,24 +35,24 @@ custom_css = """
|
|
35 |
padding-right: 5rem;
|
36 |
}
|
37 |
h1 {
|
38 |
-
color: #
|
39 |
}
|
40 |
[data-testid="stDecoration"] {
|
41 |
display: none;
|
42 |
}
|
43 |
h2 {
|
44 |
-
color: #
|
45 |
-
border-bottom: 2px solid #
|
46 |
}
|
47 |
|
48 |
h3 {
|
49 |
-
color: #
|
50 |
-
border-bottom: 2px solid #
|
51 |
}
|
52 |
|
53 |
/* Personnaliser le bouton */
|
54 |
.stButton button {
|
55 |
-
background-color: #
|
56 |
color: white !important;
|
57 |
border-radius: 5px !important;
|
58 |
border: none !important;
|
@@ -136,6 +136,8 @@ plot_label_distribution()
|
|
136 |
|
137 |
# Section Méthodologie
|
138 |
st.header('Opinion Classification')
|
|
|
|
|
139 |
st.write("""
|
140 |
Firstly, we developed and utilized a labeling script specifically designed to classify tweets into three categories: FOR, AGAINST, and NEUTRAL. In total, 500 tweets were labeled, providing a solid foundation for testing and training our opinion classification models. We adopted an 80/20 data split for training and testing, respectively, to maximize learning while effectively validating the model's performance. This relatively small dataset was chosen due to time constraints that did not allow for more extensive labeling.""")
|
141 |
st.subheader('Bert Fine-Tuning')
|
@@ -157,6 +159,8 @@ st.table(df_training)
|
|
157 |
|
158 |
# Section Présentation des Résultats
|
159 |
st.header('Results')
|
|
|
|
|
160 |
# Accuracy du modèle par classe et globale
|
161 |
data_accuracy = {
|
162 |
"Accuracies": ["Global Accuracy", "Class 0 FOR Accuracy", "Class 1 AGAINST Accuracy", "Class 2 NEUTRAL Accuracy"],
|
@@ -167,6 +171,8 @@ st.table(df_accuracy)
|
|
167 |
|
168 |
# Perspectives d'améliorations
|
169 |
st.subheader('Perspectives of Improvement')
|
|
|
|
|
170 |
st.write("""
|
171 |
The performance of our model was significantly impacted by class imbalance in the dataset, with the "against" class being underrepresented. To address this issue, we are considering two main improvement strategies:
|
172 |
""")
|
|
|
35 |
padding-right: 5rem;
|
36 |
}
|
37 |
h1 {
|
38 |
+
color: #004c75;
|
39 |
}
|
40 |
[data-testid="stDecoration"] {
|
41 |
display: none;
|
42 |
}
|
43 |
h2 {
|
44 |
+
color: #004c75;
|
45 |
+
border-bottom: 2px solid #004c75;
|
46 |
}
|
47 |
|
48 |
h3 {
|
49 |
+
color: #004c75;
|
50 |
+
border-bottom: 2px solid #004c75;
|
51 |
}
|
52 |
|
53 |
/* Personnaliser le bouton */
|
54 |
.stButton button {
|
55 |
+
background-color: #004c75 !important;
|
56 |
color: white !important;
|
57 |
border-radius: 5px !important;
|
58 |
border: none !important;
|
|
|
136 |
|
137 |
# Section Méthodologie
|
138 |
st.header('Opinion Classification')
|
139 |
+
st.text("") # Ajouter une ligne vide
|
140 |
+
|
141 |
st.write("""
|
142 |
Firstly, we developed and utilized a labeling script specifically designed to classify tweets into three categories: FOR, AGAINST, and NEUTRAL. In total, 500 tweets were labeled, providing a solid foundation for testing and training our opinion classification models. We adopted an 80/20 data split for training and testing, respectively, to maximize learning while effectively validating the model's performance. This relatively small dataset was chosen due to time constraints that did not allow for more extensive labeling.""")
|
143 |
st.subheader('Bert Fine-Tuning')
|
|
|
159 |
|
160 |
# Section Présentation des Résultats
|
161 |
st.header('Results')
|
162 |
+
st.text("") # Ajouter une ligne vide
|
163 |
+
|
164 |
# Accuracy du modèle par classe et globale
|
165 |
data_accuracy = {
|
166 |
"Accuracies": ["Global Accuracy", "Class 0 FOR Accuracy", "Class 1 AGAINST Accuracy", "Class 2 NEUTRAL Accuracy"],
|
|
|
171 |
|
172 |
# Perspectives d'améliorations
|
173 |
st.subheader('Perspectives of Improvement')
|
174 |
+
st.text("") # Ajouter une ligne vide
|
175 |
+
|
176 |
st.write("""
|
177 |
The performance of our model was significantly impacted by class imbalance in the dataset, with the "against" class being underrepresented. To address this issue, we are considering two main improvement strategies:
|
178 |
""")
|