ilsa15 commited on
Commit
a6a1010
Β·
verified Β·
1 Parent(s): fa243e5

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +251 -1
app.py CHANGED
@@ -93,4 +93,254 @@ def create_dataset():
93
  "Beautiful design and impressive performance",
94
  "I enjoyed every moment of the event",
95
  "The results are better than I expected",
96
- "Excellent quality, friendly service, and fair price",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93
  "Beautiful design and impressive performance",
94
  "I enjoyed every moment of the event",
95
  "The results are better than I expected",
96
+ "Excellent quality, friendly service, and fair price",
97
+ "This book is inspiring and well written",
98
+ "The update made the software much better",
99
+ "I am satisfied with my purchase",
100
+ "A lovely place with a welcoming atmosphere",
101
+ "The course is helpful and easy to understand",
102
+ "The new feature saves time and works well",
103
+ "The staff were polite and made us feel welcome",
104
+ "The trip was comfortable and memorable",
105
+ "The headphones sound clear and feel comfortable",
106
+ "I would happily buy this again",
107
+ ],
108
+ "negative": [
109
+ "I hate this product, it is terrible",
110
+ "The service was awful and the staff were rude",
111
+ "This was a disappointing experience",
112
+ "The movie was boring and poorly made",
113
+ "I am unhappy with the quality",
114
+ "The food tasted stale and unpleasant",
115
+ "Nothing worked and I cannot recommend it",
116
+ "The delivery was late and the package was damaged",
117
+ "The support team ignored my problem",
118
+ "I had a horrible day and feel upset",
119
+ "The app is confusing and keeps crashing",
120
+ "Poor design and disappointing performance",
121
+ "I regret attending this event",
122
+ "The results were much worse than expected",
123
+ "Bad quality, rude service, and high price",
124
+ "This book is dull and badly written",
125
+ "The update made the software worse",
126
+ "I am dissatisfied with my purchase",
127
+ "A noisy place with an unpleasant atmosphere",
128
+ "The course is confusing and not helpful",
129
+ "The new feature wastes time and fails often",
130
+ "The staff were impolite and unwelcoming",
131
+ "The trip was uncomfortable and stressful",
132
+ "The headphones sound poor and hurt my ears",
133
+ "I would never buy this again",
134
+ ],
135
+ "neutral": [
136
+ "The product arrived in a standard box",
137
+ "The service started at nine in the morning",
138
+ "I watched the movie yesterday",
139
+ "The package contains three separate items",
140
+ "The food was served on a white plate",
141
+ "The app has five tabs on the main screen",
142
+ "The delivery took two days",
143
+ "The support team replied to my message",
144
+ "I attended the event on Friday",
145
+ "The book has twelve chapters",
146
+ "The software update was released today",
147
+ "I purchased the item last week",
148
+ "The restaurant is located near the station",
149
+ "The course includes six lessons",
150
+ "The feature is available in the settings menu",
151
+ "The staff asked me to wait outside",
152
+ "The trip lasted around four hours",
153
+ "The headphones come with a charging cable",
154
+ "The meeting is scheduled for tomorrow",
155
+ "The report contains several sections",
156
+ "The website has a search bar",
157
+ "The order includes a receipt",
158
+ "The room has two windows",
159
+ "The class begins at ten o'clock",
160
+ "The device is available in black and white",
161
+ ],
162
+ }
163
+ return pd.DataFrame(
164
+ [
165
+ {"text": sentence, "sentiment": label}
166
+ for label, sentences in examples.items()
167
+ for sentence in sentences
168
+ ]
169
+ )
170
+
171
+
172
+ DATA = create_dataset()
173
+ DATA["clean_text"] = DATA["text"].apply(preprocess_text)
174
+
175
+ X_train, X_test, y_train, y_test = train_test_split(
176
+ DATA["clean_text"],
177
+ DATA["sentiment"],
178
+ test_size=0.25,
179
+ random_state=42,
180
+ stratify=DATA["sentiment"],
181
+ )
182
+
183
+ MODEL = Pipeline(
184
+ [
185
+ ("tfidf", TfidfVectorizer(ngram_range=(1, 2), sublinear_tf=True)),
186
+ ("classifier", LogisticRegression(max_iter=1000, class_weight="balanced")),
187
+ ]
188
+ )
189
+ MODEL.fit(X_train, y_train)
190
+
191
+ TEST_PREDICTIONS = MODEL.predict(X_test)
192
+ LABELS = sorted(DATA["sentiment"].unique())
193
+
194
+ ACCURACY = accuracy_score(y_test, TEST_PREDICTIONS)
195
+ MACRO_F1 = f1_score(y_test, TEST_PREDICTIONS, average="macro", zero_division=0)
196
+ MACRO_PRECISION = precision_score(
197
+ y_test, TEST_PREDICTIONS, average="macro", zero_division=0
198
+ )
199
+ MACRO_RECALL = recall_score(
200
+ y_test, TEST_PREDICTIONS, average="macro", zero_division=0
201
+ )
202
+ REPORT_DF = pd.DataFrame(
203
+ classification_report(
204
+ y_test, TEST_PREDICTIONS, output_dict=True, zero_division=0
205
+ )
206
+ ).T.round(3)
207
+ CM = confusion_matrix(y_test, TEST_PREDICTIONS, labels=LABELS)
208
+
209
+
210
+ def predict_sentiment(text):
211
+ if not text or not text.strip():
212
+ return "Please enter a sentence or review.", None, ""
213
+
214
+ cleaned = preprocess_text(text)
215
+ if not cleaned:
216
+ return (
217
+ "No useful words remained after preprocessing. Please try a longer sentence.",
218
+ None,
219
+ "",
220
+ )
221
+
222
+ prediction = MODEL.predict([cleaned])[0]
223
+ probabilities = MODEL.predict_proba([cleaned])[0]
224
+ probability_data = {
225
+ label.title(): float(probability)
226
+ for label, probability in zip(MODEL.classes_, probabilities)
227
+ }
228
+ icon = {"positive": "😊", "negative": "😞", "neutral": "😐"}.get(
229
+ prediction, "πŸ’¬"
230
+ )
231
+ result = (
232
+ f"{icon} **Predicted sentiment: {prediction.title()}**\n\n"
233
+ f"Estimated model confidence: **{probability_data[prediction.title()]:.1%}**\n\n"
234
+ "_Confidence is the model's estimate, not a guarantee._"
235
+ )
236
+ return result, probability_data, cleaned
237
+
238
+
239
+ def make_confusion_matrix():
240
+ fig, ax = plt.subplots(figsize=(6, 4))
241
+ image = ax.imshow(CM, interpolation="nearest", cmap="Blues")
242
+ fig.colorbar(image, ax=ax)
243
+ ax.set(
244
+ xticks=range(len(LABELS)),
245
+ yticks=range(len(LABELS)),
246
+ xticklabels=[label.title() for label in LABELS],
247
+ yticklabels=[label.title() for label in LABELS],
248
+ xlabel="Predicted label",
249
+ ylabel="True label",
250
+ title="Confusion Matrix",
251
+ )
252
+ plt.setp(ax.get_xticklabels(), rotation=20, ha="right", rotation_mode="anchor")
253
+ threshold = CM.max() / 2 if CM.size else 0
254
+ for row in range(CM.shape[0]):
255
+ for col in range(CM.shape[1]):
256
+ ax.text(
257
+ col,
258
+ row,
259
+ str(CM[row, col]),
260
+ ha="center",
261
+ va="center",
262
+ color="white" if CM[row, col] > threshold else "black",
263
+ )
264
+ fig.tight_layout()
265
+ return fig
266
+
267
+
268
+ def get_dataset_csv():
269
+ return DATA[["text", "sentiment"]].to_csv(index=False).encode("utf-8")
270
+
271
+
272
+ with gr.Blocks(title="Multi-Class Sentiment Classifier", theme=gr.themes.Soft()) as demo:
273
+ gr.Markdown(
274
+ """
275
+ # πŸ’¬ Intelligent Multi-Class Sentiment Classifier
276
+
277
+ Classify text as **positive**, **negative**, or **neutral** using
278
+ text preprocessing, TF-IDF features, and a Scikit-learn Logistic
279
+ Regression model.
280
+ """
281
+ )
282
+
283
+ with gr.Tab("πŸ”Ž Predict Sentiment"):
284
+ text_input = gr.Textbox(
285
+ label="Enter a sentence or review",
286
+ placeholder="Example: The product is excellent and easy to use!",
287
+ lines=4,
288
+ )
289
+ analyze_button = gr.Button("Analyze Sentiment", variant="primary")
290
+ prediction_output = gr.Markdown()
291
+ probability_output = gr.Label(label="Class probabilities", num_top_classes=3)
292
+ cleaned_output = gr.Textbox(label="Preprocessed text", interactive=False)
293
+ analyze_button.click(
294
+ fn=predict_sentiment,
295
+ inputs=text_input,
296
+ outputs=[prediction_output, probability_output, cleaned_output],
297
+ )
298
+ text_input.submit(
299
+ fn=predict_sentiment,
300
+ inputs=text_input,
301
+ outputs=[prediction_output, probability_output, cleaned_output],
302
+ )
303
+ gr.Examples(
304
+ examples=[
305
+ ["I absolutely love this product, it works wonderfully"],
306
+ ["The service was awful and the staff were rude"],
307
+ ["The meeting is scheduled for tomorrow"],
308
+ ],
309
+ inputs=text_input,
310
+ )
311
+
312
+ with gr.Tab("πŸ“Š Model Evaluation"):
313
+ gr.Markdown(
314
+ "### Performance on a held-out test set\n"
315
+ "These metrics are calculated from a stratified split of the built-in sample dataset."
316
+ )
317
+ with gr.Row():
318
+ gr.Number(value=round(ACCURACY, 3), label="Accuracy", interactive=False)
319
+ gr.Number(value=round(MACRO_F1, 3), label="Macro F1-score", interactive=False)
320
+ gr.Number(value=round(MACRO_PRECISION, 3), label="Macro precision", interactive=False)
321
+ gr.Number(value=round(MACRO_RECALL, 3), label="Macro recall", interactive=False)
322
+ gr.Markdown("### Classification report")
323
+ gr.Dataframe(value=REPORT_DF.reset_index(names=["class"]), interactive=False)
324
+ gr.Markdown("### Confusion matrix")
325
+ gr.Plot(value=make_confusion_matrix())
326
+
327
+ with gr.Tab("πŸ—‚οΈ Dataset"):
328
+ gr.Markdown(
329
+ f"The built-in sample dataset contains **{len(DATA)} text examples** "
330
+ "across three sentiment classes. It is intended for coursework demonstration."
331
+ )
332
+ gr.Dataframe(value=DATA[["text", "sentiment"]], interactive=False)
333
+ gr.File(
334
+ value=("sentiment_dataset.csv", get_dataset_csv()),
335
+ label="Sample dataset CSV",
336
+ interactive=False,
337
+ )
338
+
339
+ gr.Markdown(
340
+ "---\n"
341
+ "**Note:** This is an educational prototype trained on a small sample dataset. "
342
+ "Its predictions and scores should not be treated as production-grade results."
343
+ )
344
+
345
+ if __name__ == "__main__":
346
+ demo.launch()