YixuanWang
commited on
Commit
•
37190a8
1
Parent(s):
596f852
Update app.py
Browse files
app.py
CHANGED
@@ -1,13 +1,13 @@
|
|
1 |
import gradio as gr
|
2 |
import pandas as pd
|
3 |
import numpy as np
|
|
|
|
|
4 |
from textblob import TextBlob
|
5 |
from typing import List, Dict, Tuple
|
6 |
from dataclasses import dataclass
|
7 |
from pathlib import Path
|
8 |
import logging
|
9 |
-
import re
|
10 |
-
from datetime import datetime
|
11 |
|
12 |
logging.basicConfig(level=logging.INFO)
|
13 |
logger = logging.getLogger(__name__)
|
@@ -20,12 +20,18 @@ class RecommendationWeights:
|
|
20 |
|
21 |
class TweetPreprocessor:
|
22 |
def __init__(self, data_path: Path):
|
23 |
-
"""Initialize the preprocessor with data path."""
|
24 |
self.data = self._load_data(data_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
@staticmethod
|
27 |
def _load_data(data_path: Path) -> pd.DataFrame:
|
28 |
-
"""Load and validate the dataset."""
|
29 |
try:
|
30 |
data = pd.read_csv(data_path)
|
31 |
required_columns = {'Text', 'Retweets', 'Likes'}
|
@@ -36,39 +42,28 @@ class TweetPreprocessor:
|
|
36 |
logger.error(f"Error loading data: {e}")
|
37 |
raise
|
38 |
|
39 |
-
def _clean_text(self, text: str) -> str:
|
40 |
-
"""Clean text content."""
|
41 |
-
if pd.isna(text) or len(str(text).strip()) < 10:
|
42 |
-
return ""
|
43 |
-
|
44 |
-
text = re.sub(r'http\S+|www.\S+', '', str(text))
|
45 |
-
text = re.sub(r'[^\w\s]', '', text)
|
46 |
-
text = ' '.join(text.split())
|
47 |
-
return text
|
48 |
-
|
49 |
def calculate_metrics(self) -> pd.DataFrame:
|
50 |
-
|
51 |
-
self.data['
|
52 |
-
self.data = self.data[self.data['Clean_Text'].str.len() > 0]
|
53 |
|
54 |
-
|
55 |
-
self.data['Popularity'] = self.
|
|
|
|
|
56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
return self.data
|
58 |
-
|
59 |
-
@staticmethod
|
60 |
-
def _get_sentiment(text: str) -> float:
|
61 |
-
"""Calculate sentiment polarity for a text."""
|
62 |
-
try:
|
63 |
-
return TextBlob(str(text)).sentiment.polarity
|
64 |
-
except Exception as e:
|
65 |
-
logger.warning(f"Error calculating sentiment: {e}")
|
66 |
-
return 0.0
|
67 |
-
|
68 |
-
def _normalize_popularity(self) -> pd.Series:
|
69 |
-
"""Normalize popularity scores."""
|
70 |
-
popularity = self.data['Retweets'] + self.data['Likes']
|
71 |
-
return (popularity - popularity.min()) / (popularity.max() - popularity.min() + 1e-6)
|
72 |
|
73 |
class RecommendationSystem:
|
74 |
def __init__(self, data_path: Path):
|
@@ -77,36 +72,28 @@ class RecommendationSystem:
|
|
77 |
self.setup_system()
|
78 |
|
79 |
def setup_system(self):
|
80 |
-
"""Initialize the system with preprocessed data."""
|
81 |
self.data = self.preprocessor.calculate_metrics()
|
82 |
|
83 |
-
def
|
84 |
-
|
|
|
|
|
85 |
normalized_weights = self._normalize_weights(weights)
|
86 |
|
87 |
-
self.data['Credibility'] = np.random.choice([0, 1], size=len(self.data), p=[0.3, 0.7])
|
88 |
-
|
89 |
self.data['Final_Score'] = (
|
90 |
self.data['Credibility'] * normalized_weights.visibility +
|
91 |
self.data['Sentiment'] * normalized_weights.sentiment +
|
92 |
self.data['Popularity'] * normalized_weights.popularity
|
93 |
)
|
94 |
|
95 |
-
def get_recommendations(self, weights: RecommendationWeights, num_recommendations: int = 10) -> Dict:
|
96 |
-
"""Get tweet recommendations based on weights."""
|
97 |
-
if not self._validate_weights(weights):
|
98 |
-
return {"error": "Invalid weights provided"}
|
99 |
-
|
100 |
-
self.recalculate_scores(weights)
|
101 |
-
|
102 |
top_recommendations = (
|
103 |
-
self.data.nlargest(
|
|
|
104 |
)
|
105 |
|
106 |
return self._format_recommendations(top_recommendations)
|
107 |
|
108 |
def _format_recommendations(self, recommendations: pd.DataFrame) -> Dict:
|
109 |
-
"""Format recommendations for display."""
|
110 |
formatted_results = []
|
111 |
for _, row in recommendations.iterrows():
|
112 |
score_details = {
|
@@ -118,7 +105,7 @@ class RecommendationSystem:
|
|
118 |
}
|
119 |
|
120 |
formatted_results.append({
|
121 |
-
"text": row['
|
122 |
"scores": score_details
|
123 |
})
|
124 |
|
@@ -129,7 +116,6 @@ class RecommendationSystem:
|
|
129 |
|
130 |
@staticmethod
|
131 |
def _get_sentiment_label(sentiment_score: float) -> str:
|
132 |
-
"""Convert sentiment score to label."""
|
133 |
if sentiment_score > 0.3:
|
134 |
return "Positive"
|
135 |
elif sentiment_score < -0.3:
|
@@ -138,12 +124,10 @@ class RecommendationSystem:
|
|
138 |
|
139 |
@staticmethod
|
140 |
def _validate_weights(weights: RecommendationWeights) -> bool:
|
141 |
-
"""Validate that weights are non-negative."""
|
142 |
return all(getattr(weights, field) >= 0 for field in weights.__dataclass_fields__)
|
143 |
|
144 |
@staticmethod
|
145 |
def _normalize_weights(weights: RecommendationWeights) -> RecommendationWeights:
|
146 |
-
"""Normalize weights to sum to 1."""
|
147 |
total = weights.visibility + weights.sentiment + weights.popularity
|
148 |
if total == 0:
|
149 |
return RecommendationWeights(1/3, 1/3, 1/3)
|
@@ -155,7 +139,6 @@ class RecommendationSystem:
|
|
155 |
|
156 |
@staticmethod
|
157 |
def _get_score_explanation() -> Dict[str, str]:
|
158 |
-
"""Provide explanation for different score components."""
|
159 |
return {
|
160 |
"Credibility": "Content reliability assessment",
|
161 |
"Sentiment": "Text emotional analysis result",
|
@@ -163,7 +146,6 @@ class RecommendationSystem:
|
|
163 |
}
|
164 |
|
165 |
def create_gradio_interface(recommendation_system: RecommendationSystem) -> gr.Interface:
|
166 |
-
"""Create and configure the Gradio interface."""
|
167 |
with gr.Blocks(theme=gr.themes.Soft()) as interface:
|
168 |
gr.Markdown("""
|
169 |
# Tweet Recommendation System
|
@@ -224,7 +206,6 @@ def create_gradio_interface(recommendation_system: RecommendationSystem) -> gr.I
|
|
224 |
return html
|
225 |
|
226 |
def get_recommendations_with_weights(v, s, p):
|
227 |
-
"""Get recommendations with current weights."""
|
228 |
weights = RecommendationWeights(v, s, p)
|
229 |
return format_recommendations(recommendation_system.get_recommendations(weights))
|
230 |
|
@@ -237,7 +218,6 @@ def create_gradio_interface(recommendation_system: RecommendationSystem) -> gr.I
|
|
237 |
return interface
|
238 |
|
239 |
def main():
|
240 |
-
"""Main function to run the application."""
|
241 |
try:
|
242 |
recommendation_system = RecommendationSystem(
|
243 |
data_path=Path('twitter_dataset.csv')
|
|
|
1 |
import gradio as gr
|
2 |
import pandas as pd
|
3 |
import numpy as np
|
4 |
+
import torch
|
5 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
6 |
from textblob import TextBlob
|
7 |
from typing import List, Dict, Tuple
|
8 |
from dataclasses import dataclass
|
9 |
from pathlib import Path
|
10 |
import logging
|
|
|
|
|
11 |
|
12 |
logging.basicConfig(level=logging.INFO)
|
13 |
logger = logging.getLogger(__name__)
|
|
|
20 |
|
21 |
class TweetPreprocessor:
|
22 |
def __init__(self, data_path: Path):
|
|
|
23 |
self.data = self._load_data(data_path)
|
24 |
+
self.model_name = "hamzab/roberta-fake-news-classification"
|
25 |
+
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
26 |
+
self.model, self.tokenizer = self._load_model()
|
27 |
+
|
28 |
+
def _load_model(self):
|
29 |
+
tokenizer = AutoTokenizer.from_pretrained(self.model_name)
|
30 |
+
model = AutoModelForSequenceClassification.from_pretrained(self.model_name).to(self.device)
|
31 |
+
return model, tokenizer
|
32 |
|
33 |
@staticmethod
|
34 |
def _load_data(data_path: Path) -> pd.DataFrame:
|
|
|
35 |
try:
|
36 |
data = pd.read_csv(data_path)
|
37 |
required_columns = {'Text', 'Retweets', 'Likes'}
|
|
|
42 |
logger.error(f"Error loading data: {e}")
|
43 |
raise
|
44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
def calculate_metrics(self) -> pd.DataFrame:
|
46 |
+
# Calculate sentiment
|
47 |
+
self.data['Sentiment'] = self.data['Text'].apply(lambda x: TextBlob(x).sentiment.polarity)
|
|
|
48 |
|
49 |
+
# Calculate popularity
|
50 |
+
self.data['Popularity'] = self.data['Retweets'] + self.data['Likes']
|
51 |
+
self.data['Popularity'] = (self.data['Popularity'] - self.data['Popularity'].mean()) / self.data['Popularity'].std()
|
52 |
+
self.data['Popularity'] = self.data['Popularity'] / self.data['Popularity'].abs().max()
|
53 |
|
54 |
+
# Calculate credibility using fake news model
|
55 |
+
batch_size = 100
|
56 |
+
predictions = []
|
57 |
+
for i in range(0, len(self.data), batch_size):
|
58 |
+
batch = self.data['Text'][i:i + batch_size].tolist()
|
59 |
+
inputs = self.tokenizer(batch, return_tensors="pt", padding=True, truncation=True, max_length=128)
|
60 |
+
inputs = {key: val.to(self.device) for key, val in inputs.items()}
|
61 |
+
with torch.no_grad():
|
62 |
+
outputs = self.model(**inputs)
|
63 |
+
predictions.extend(outputs.logits.argmax(dim=1).cpu().numpy())
|
64 |
+
|
65 |
+
self.data['Credibility'] = [1 if pred == 1 else -1 for pred in predictions]
|
66 |
return self.data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
67 |
|
68 |
class RecommendationSystem:
|
69 |
def __init__(self, data_path: Path):
|
|
|
72 |
self.setup_system()
|
73 |
|
74 |
def setup_system(self):
|
|
|
75 |
self.data = self.preprocessor.calculate_metrics()
|
76 |
|
77 |
+
def get_recommendations(self, weights: RecommendationWeights, num_recommendations: int = 10) -> Dict:
|
78 |
+
if not self._validate_weights(weights):
|
79 |
+
return {"error": "Invalid weights provided"}
|
80 |
+
|
81 |
normalized_weights = self._normalize_weights(weights)
|
82 |
|
|
|
|
|
83 |
self.data['Final_Score'] = (
|
84 |
self.data['Credibility'] * normalized_weights.visibility +
|
85 |
self.data['Sentiment'] * normalized_weights.sentiment +
|
86 |
self.data['Popularity'] * normalized_weights.popularity
|
87 |
)
|
88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
top_recommendations = (
|
90 |
+
self.data.nlargest(100, 'Final_Score')
|
91 |
+
.sample(num_recommendations)
|
92 |
)
|
93 |
|
94 |
return self._format_recommendations(top_recommendations)
|
95 |
|
96 |
def _format_recommendations(self, recommendations: pd.DataFrame) -> Dict:
|
|
|
97 |
formatted_results = []
|
98 |
for _, row in recommendations.iterrows():
|
99 |
score_details = {
|
|
|
105 |
}
|
106 |
|
107 |
formatted_results.append({
|
108 |
+
"text": row['Text'],
|
109 |
"scores": score_details
|
110 |
})
|
111 |
|
|
|
116 |
|
117 |
@staticmethod
|
118 |
def _get_sentiment_label(sentiment_score: float) -> str:
|
|
|
119 |
if sentiment_score > 0.3:
|
120 |
return "Positive"
|
121 |
elif sentiment_score < -0.3:
|
|
|
124 |
|
125 |
@staticmethod
|
126 |
def _validate_weights(weights: RecommendationWeights) -> bool:
|
|
|
127 |
return all(getattr(weights, field) >= 0 for field in weights.__dataclass_fields__)
|
128 |
|
129 |
@staticmethod
|
130 |
def _normalize_weights(weights: RecommendationWeights) -> RecommendationWeights:
|
|
|
131 |
total = weights.visibility + weights.sentiment + weights.popularity
|
132 |
if total == 0:
|
133 |
return RecommendationWeights(1/3, 1/3, 1/3)
|
|
|
139 |
|
140 |
@staticmethod
|
141 |
def _get_score_explanation() -> Dict[str, str]:
|
|
|
142 |
return {
|
143 |
"Credibility": "Content reliability assessment",
|
144 |
"Sentiment": "Text emotional analysis result",
|
|
|
146 |
}
|
147 |
|
148 |
def create_gradio_interface(recommendation_system: RecommendationSystem) -> gr.Interface:
|
|
|
149 |
with gr.Blocks(theme=gr.themes.Soft()) as interface:
|
150 |
gr.Markdown("""
|
151 |
# Tweet Recommendation System
|
|
|
206 |
return html
|
207 |
|
208 |
def get_recommendations_with_weights(v, s, p):
|
|
|
209 |
weights = RecommendationWeights(v, s, p)
|
210 |
return format_recommendations(recommendation_system.get_recommendations(weights))
|
211 |
|
|
|
218 |
return interface
|
219 |
|
220 |
def main():
|
|
|
221 |
try:
|
222 |
recommendation_system = RecommendationSystem(
|
223 |
data_path=Path('twitter_dataset.csv')
|