Jan90's picture
Update app.py
6100bab verified
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModelForSequenceClassification, TextClassificationPipeline
import torch
import gradio as gr
from openpyxl import load_workbook
from numpy import mean
import pandas as pd
import matplotlib.pyplot as plt
theme = gr.themes.Soft(
primary_hue="amber",
secondary_hue="amber",
neutral_hue="stone",
)
# Load tokenizers and models
tokenizer = AutoTokenizer.from_pretrained("suriya7/bart-finetuned-text-summarization")
model = AutoModelForSeq2SeqLM.from_pretrained("suriya7/bart-finetuned-text-summarization")
tokenizer_keywords = AutoTokenizer.from_pretrained("transformer3/H2-keywordextractor")
model_keywords = AutoModelForSeq2SeqLM.from_pretrained("transformer3/H2-keywordextractor")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
new_model = AutoModelForSequenceClassification.from_pretrained('roberta-rating')
new_tokenizer = AutoTokenizer.from_pretrained('roberta-rating')
classifier = TextClassificationPipeline(model=new_model, tokenizer=new_tokenizer, device=device)
label_mapping = {1: '1/5', 2: '2/5', 3: '3/5', 4: '4/5', 5: '5/5'}
# Function to display and filter the Excel workbook
def filter_xl(file, keywords):
# Load the workbook and convert it to a DataFrame
workbook = load_workbook(filename=file)
sheet = workbook.active
data = sheet.values
columns = next(data)[0:]
df = pd.DataFrame(data, columns=columns)
if keywords:
keyword_list = keywords.split(',')
for keyword in keyword_list:
df = df[df.apply(lambda row: row.astype(str).str.contains(keyword.strip(), case=False).any(), axis=1)]
return df
# Function to calculate overall rating from filtered data
def calculate_rating(filtered_df):
reviews = filtered_df.to_numpy().flatten()
ratings = []
for review in reviews:
if pd.notna(review):
rating = int(classifier(review)[0]['label'].split('_')[1])
ratings.append(rating)
return round(mean(ratings), 2), ratings
# Function to calculate results including summary, keywords, and sentiment
def calculate_results(file, keywords):
filtered_df = filter_xl(file, keywords)
overall_rating, ratings = calculate_rating(filtered_df)
# Summarize and extract keywords from the filtered reviews
text = " ".join(filtered_df.to_numpy().flatten())
inputs = tokenizer([text], max_length=1024, truncation=True, return_tensors="pt")
summary_ids = model.generate(inputs["input_ids"], num_beams=2, min_length=10, max_length=50)
summary = tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
summary = summary.replace("I", "They").replace("my", "their").replace("me", "them")
inputs_keywords = tokenizer_keywords([text], max_length=1024, truncation=True, return_tensors="pt")
summary_ids_keywords = model_keywords.generate(inputs_keywords["input_ids"], num_beams=2, min_length=0, max_length=100)
keywords = tokenizer_keywords.batch_decode(summary_ids_keywords, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
# Determine overall sentiment
sentiments = []
for review in filtered_df.to_numpy().flatten():
if pd.notna(review):
sentiment = classifier(review)[0]['label']
sentiment_label = "Positive" if sentiment == "LABEL_4" or sentiment == "LABEL_5" else "Negative" if sentiment == "LABEL_1" or sentiment == "LABEL_2" else "Neutral"
sentiments.append(sentiment_label)
overall_sentiment = "Positive" if sentiments.count("Positive") > sentiments.count("Negative") else "Negative" if sentiments.count("Negative") > sentiments.count("Positive") else "Neutral"
return overall_rating, summary, keywords, overall_sentiment, ratings, sentiments
# Function to analyze a single review
def analyze_review(review):
if not review.strip():
return "Error: No text provided", "Error: No text provided", "Error: No text provided", "Error: No text provided"
# Calculate rating
rating = int(classifier(review)[0]['label'].split('_')[1])
# Summarize review
inputs = tokenizer([review], max_length=1024, truncation=True, return_tensors="pt")
summary_ids = model.generate(inputs["input_ids"], num_beams=2, min_length=10, max_length=50)
summary = tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
summary = summary.replace("I", "he/she").replace("my", "his/her").replace("me", "him/her")
# Extract keywords
inputs_keywords = tokenizer_keywords([review], max_length=1024, truncation=True, return_tensors="pt")
summary_ids_keywords = model_keywords.generate(inputs_keywords["input_ids"], num_beams=2, min_length=0, max_length=100)
keywords = tokenizer_keywords.batch_decode(summary_ids_keywords, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
# Determine sentiment
sentiment = classifier(review)[0]['label']
sentiment_label = "Positive" if sentiment == "LABEL_4" or sentiment == "LABEL_5" else "Negative" if sentiment == "LABEL_1" or sentiment == "LABEL_2" else "Neutral"
return rating, summary, keywords, sentiment_label
# Function to count rows in the filtered DataFrame
def count_rows(filtered_df):
return len(filtered_df)
# Function to plot ratings
def plot_ratings(ratings):
plt.figure(figsize=(10, 5))
plt.hist(ratings, bins=range(1, 7), edgecolor='black', align='left')
plt.xlabel('Rating')
plt.ylabel('Frequency')
plt.title('Distribution of Ratings')
plt.xticks(range(1, 6))
plt.grid(True)
plt.savefig('ratings_distribution.png')
return 'ratings_distribution.png'
# Function to plot sentiments
def plot_sentiments(sentiments):
sentiment_counts = pd.Series(sentiments).value_counts()
plt.figure(figsize=(10, 5))
sentiment_counts.plot(kind='bar', color=['green', 'red', 'blue'])
plt.xlabel('Sentiment')
plt.ylabel('Frequency')
plt.title('Distribution of Sentiments')
plt.grid(True)
plt.savefig('sentiments_distribution.png')
return 'sentiments_distribution.png'
# Gradio interface
with gr.Blocks(theme=theme) as demo:
gr.Markdown("<h1 style='text-align: center;'>Feedback and Auditing Survey AI Analyzer</h1><br>")
with gr.Tabs():
with gr.TabItem("Upload and Filter"):
with gr.Row():
with gr.Column(scale=1):
excel_file = gr.File(label="Upload Excel File")
#excel_file = gr.File(label="Upload Excel File", file_types=[".xlsx", ".xlsm", ".xltx", ".xltm"])
keywords_input = gr.Textbox(label="Filter by Keywords (comma-separated)")
display_button = gr.Button("Display and Filter Excel Data")
clear_button_upload = gr.Button("Clear")
row_count = gr.Textbox(label="Number of Rows", interactive=False)
with gr.Column(scale=3):
filtered_data = gr.Dataframe(label="Filtered Excel Contents")
with gr.TabItem("Calculate Results"):
with gr.Row():
with gr.Column():
overall_rating = gr.Textbox(label="Overall Rating")
summary = gr.Textbox(label="Summary")
keywords_output = gr.Textbox(label="Keywords")
overall_sentiment = gr.Textbox(label="Overall Sentiment")
calculate_button = gr.Button("Calculate Results")
with gr.Column():
ratings_graph = gr.Image(label="Ratings Distribution")
sentiments_graph = gr.Image(label="Sentiments Distribution")
calculate_graph_button = gr.Button("Calculate Graph Results")
with gr.TabItem("Testing Area / Write a Review"):
with gr.Row():
with gr.Column(scale=2):
review_input = gr.Textbox(label="Write your review here")
analyze_button = gr.Button("Analyze Review")
clear_button_review = gr.Button("Clear")
with gr.Column(scale=2):
review_rating = gr.Textbox(label="Rating")
review_summary = gr.Textbox(label="Summary")
review_keywords = gr.Textbox(label="Keywords")
review_sentiment = gr.Textbox(label="Sentiment")
display_button.click(lambda file, keywords: (filter_xl(file, keywords), count_rows(filter_xl(file, keywords))), inputs=[excel_file, keywords_input], outputs=[filtered_data, row_count])
calculate_graph_button.click(lambda file, keywords: (*calculate_results(file, keywords)[:4], plot_ratings(calculate_results(file, keywords)[4]), plot_sentiments(calculate_results(file, keywords)[5])), inputs=[excel_file, keywords_input], outputs=[overall_rating, summary, keywords_output, overall_sentiment, ratings_graph, sentiments_graph])
calculate_button.click(lambda file, keywords: (*calculate_results(file, keywords)[:4], plot_ratings(calculate_results(file, keywords)[4])), inputs=[excel_file, keywords_input], outputs=[overall_rating, summary, keywords_output, overall_sentiment])
analyze_button.click(analyze_review, inputs=review_input, outputs=[review_rating, review_summary, review_keywords, review_sentiment])
clear_button_upload.click(lambda: (""), outputs=[keywords_input])
clear_button_review.click(lambda: ("", "", "", "", ""), outputs=[review_input, review_rating, review_summary, review_keywords, review_sentiment])
demo.launch(share=True)