Spaces:
Running
Running
File size: 9,095 Bytes
cc8348d bdb3169 457d325 cc8348d bdb3169 e4627b7 bdb3169 e4627b7 bdb3169 457d325 cc8348d bdb3169 457d325 cc8348d e4627b7 cc8348d e4627b7 cc8348d e4627b7 bdb3169 457d325 cc8348d e4627b7 457d325 e4627b7 457d325 e4627b7 cc8348d 457d325 cc8348d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 |
import gradio as gr
from transformers import (
DistilBertTokenizerFast,
DistilBertForSequenceClassification,
AutoTokenizer,
AutoModelForSequenceClassification,
)
from huggingface_hub import hf_hub_download
import torch
import pickle
import numpy as np
# Load models and tokenizers
models = {
"DistilBERT": {
"tokenizer": DistilBertTokenizerFast.from_pretrained("nhull/distilbert-sentiment-model"),
"model": DistilBertForSequenceClassification.from_pretrained("nhull/distilbert-sentiment-model"),
},
"Logistic Regression": {}, # Placeholder for logistic regression
"BERT Multilingual (NLP Town)": {
"tokenizer": AutoTokenizer.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
"model": AutoModelForSequenceClassification.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
},
"TinyBERT": {
"tokenizer": AutoTokenizer.from_pretrained("elo4/TinyBERT-sentiment-model"),
"model": AutoModelForSequenceClassification.from_pretrained("elo4/TinyBERT-sentiment-model"),
},
"RoBERTa": {
"tokenizer": AutoTokenizer.from_pretrained("ordek899/roberta_1to5rating_pred_for_restaur_trained_on_hotels"),
"model": AutoModelForSequenceClassification.from_pretrained("ordek899/roberta_1to5rating_pred_for_restaur_trained_on_hotels"),
}
}
# Load logistic regression model and vectorizer
logistic_regression_repo = "nhull/logistic-regression-model"
# Download and load logistic regression model
log_reg_model_path = hf_hub_download(repo_id=logistic_regression_repo, filename="logistic_regression_model.pkl")
with open(log_reg_model_path, "rb") as model_file:
log_reg_model = pickle.load(model_file)
# Download and load TF-IDF vectorizer
vectorizer_path = hf_hub_download(repo_id=logistic_regression_repo, filename="tfidf_vectorizer.pkl")
with open(vectorizer_path, "rb") as vectorizer_file:
vectorizer = pickle.load(vectorizer_file)
# Move HuggingFace models to device (if GPU is available)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
for model_data in models.values():
if "model" in model_data:
model_data["model"].to(device)
# Functions for prediction
def predict_with_distilbert(text):
tokenizer = models["DistilBERT"]["tokenizer"]
model = models["DistilBERT"]["model"]
encodings = tokenizer([text], padding=True, truncation=True, max_length=512, return_tensors="pt").to(device)
with torch.no_grad():
outputs = model(**encodings)
logits = outputs.logits
predictions = logits.argmax(axis=-1).cpu().numpy()
return int(predictions[0] + 1)
def predict_with_logistic_regression(text):
transformed_text = vectorizer.transform([text])
predictions = log_reg_model.predict(transformed_text)
return int(predictions[0])
def predict_with_bert_multilingual(text):
tokenizer = models["BERT Multilingual (NLP Town)"]["tokenizer"]
model = models["BERT Multilingual (NLP Town)"]["model"]
encodings = tokenizer([text], padding=True, truncation=True, max_length=512, return_tensors="pt").to(device)
with torch.no_grad():
outputs = model(**encodings)
logits = outputs.logits
predictions = logits.argmax(axis=-1).cpu().numpy()
return int(predictions[0] + 1)
def predict_with_tinybert(text):
tokenizer = models["TinyBERT"]["tokenizer"]
model = models["TinyBERT"]["model"]
encodings = tokenizer([text], padding=True, truncation=True, max_length=512, return_tensors="pt").to(device)
with torch.no_grad():
outputs = model(**encodings)
logits = outputs.logits
predictions = logits.argmax(axis=-1).cpu().numpy()
return int(predictions[0] + 1)
def predict_with_roberta_ordek899(text):
tokenizer = models["RoBERTa"]["tokenizer"]
model = models["RoBERTa"]["model"]
encodings = tokenizer([text], padding=True, truncation=True, max_length=512, return_tensors="pt").to(device)
with torch.no_grad():
outputs = model(**encodings)
logits = outputs.logits
predictions = logits.argmax(axis=-1).cpu().numpy()
return int(predictions[0] + 1)
# Unified function for sentiment analysis and statistics
def analyze_sentiment_and_statistics(text):
results = {
"DistilBERT": predict_with_distilbert(text),
"Logistic Regression": predict_with_logistic_regression(text),
"BERT Multilingual (NLP Town)": predict_with_bert_multilingual(text),
"TinyBERT": predict_with_tinybert(text),
"RoBERTa": predict_with_roberta_ordek899(text),
}
# Calculate statistics
scores = list(results.values())
if all(score == scores[0] for score in scores): # Check if all predictions are the same
statistics = {
"Message": "All models predict the same score.",
"Average Score": f"{scores[0]:.2f}",
}
else:
min_score_model = min(results, key=results.get)
max_score_model = max(results, key=results.get)
average_score = np.mean(scores)
statistics = {
"Lowest Score": f"{results[min_score_model]} (Model: {min_score_model})",
"Highest Score": f"{results[max_score_model]} (Model: {max_score_model})",
"Average Score": f"{average_score:.2f}",
}
return results, statistics
# Gradio Interface
with gr.Blocks(css=".gradio-container { max-width: 900px; margin: auto; padding: 20px; }") as demo:
gr.Markdown("# Sentiment Analysis App")
gr.Markdown(
"This app predicts the sentiment of the input text on a scale from 1 to 5 using multiple models and provides basic statistics."
)
with gr.Row():
with gr.Column():
text_input = gr.Textbox(
label="Enter your text here:",
lines=3,
placeholder="Type your hotel/restaurant review here..."
)
sample_reviews = [
"The hotel was fantastic! Clean rooms and excellent service.",
"The food was horrible, and the staff was rude.",
"Amazing experience overall. Highly recommend!",
"It was okay, not great but not terrible either.",
"Terrible! The room was dirty, and the service was non-existent."
]
sample_dropdown = gr.Dropdown(
choices=sample_reviews,
label="Or select a sample review:",
interactive=True
)
# Sync dropdown with text input
def update_textbox(selected_sample):
return selected_sample
sample_dropdown.change(
update_textbox,
inputs=[sample_dropdown],
outputs=[text_input]
)
with gr.Column():
analyze_button = gr.Button("Analyze Sentiment")
with gr.Row():
with gr.Column():
distilbert_output = gr.Textbox(label="Predicted Sentiment (DistilBERT)", interactive=False)
log_reg_output = gr.Textbox(label="Predicted Sentiment (Logistic Regression)", interactive=False)
bert_output = gr.Textbox(label="Predicted Sentiment (BERT Multilingual)", interactive=False)
tinybert_output = gr.Textbox(label="Predicted Sentiment (TinyBERT)", interactive=False)
roberta_ordek_output = gr.Textbox(label="Predicted Sentiment (RoBERTa)", interactive=False)
with gr.Column():
statistics_output = gr.Textbox(label="Statistics (Lowest, Highest, Average)", interactive=False)
# Button to analyze sentiment and show statistics
def process_input_and_analyze(text_input):
results, statistics = analyze_sentiment_and_statistics(text_input)
if "Message" in statistics: # All models predicted the same score
return (
f"{results['DistilBERT']}",
f"{results['Logistic Regression']}",
f"{results['BERT Multilingual (NLP Town)']}",
f"{results['TinyBERT']}",
f"{results['RoBERTa']}",
f"Statistics:\n{statistics['Message']}\nAverage Score: {statistics['Average Score']}"
)
else: # Min and Max scores are present
return (
f"{results['DistilBERT']}",
f"{results['Logistic Regression']}",
f"{results['BERT Multilingual (NLP Town)']}",
f"{results['TinyBERT']}",
f"{results['RoBERTa']}",
f"Statistics:\n{statistics['Lowest Score']}\n{statistics['Highest Score']}\nAverage Score: {statistics['Average Score']}"
)
analyze_button.click(
process_input_and_analyze,
inputs=[text_input],
outputs=[
distilbert_output,
log_reg_output,
bert_output,
tinybert_output,
roberta_ordek_output,
statistics_output
]
)
# Launch the app
demo.launch()
|