app.py

import gradio as gr
import pandas as pd



def data_pre_processing(file_responses):
    # Financial Weights are in per decas and NOT per cents
    
    ### GPT: Assuming 'Your financial allocation for Problem (in $)' column contains numerical values
    
    file_responses['''Your financial allocation for Problem 1:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a specific solution for your 1st problem.'''] = pd.to_numeric(file_responses['''Your financial allocation for Problem 1:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a specific solution for your 1st problem.'''], errors='coerce').fillna(0)
    
    file_responses['''Your financial allocation for Problem 2:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 2nd problem.'''] = pd.to_numeric(file_responses['''Your financial allocation for Problem 2:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 2nd problem.'''], errors='coerce').fillna(0)
    
    file_responses['''Your financial allocation for Problem 3:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 3rd problem.'''] = pd.to_numeric(file_responses['''Your financial allocation for Problem 3:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 3rd problem.'''], errors='coerce').fillna(0)    
    
    file_responses['''How much was your latest Tax payment (in U$D) ?
    
    Please try to be as accurate as possible:
    Eg.: If your last tax amount was INR 25,785/-; then convert it in U$D and enter only the amount as: 310.
    
    If you have never paid tax, consider putting in a realistic donation amount which wish to contribute towards helping yourself obtain the desired relief.'''
    ] = pd.to_numeric(file_responses['''How much was your latest Tax payment (in U$D) ?
    
    Please try to be as accurate as possible:
    Eg.: If your last tax amount was INR 25,785/-; then convert it in U$D and enter only the amount as: 310.
    
    If you have never paid tax, consider putting in a realistic donation amount which wish to contribute towards helping yourself obtain the desired relief.'''
    ], errors='coerce').fillna(0)
    
    
    
    
    
    # Adding a new column 'Total Allocation' by summing specific columns by their names
    file_responses['Total Allocation'] = file_responses[['''Your financial allocation for Problem 1:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a specific solution for your 1st problem.''' , '''Your financial allocation for Problem 2:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 2nd problem.''' , '''Your financial allocation for Problem 3:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 3rd problem.''']].apply(lambda x: x.clip(lower=10)).sum(axis=1)
    
    
    
    
    
    # Creating 'Financial Weight' column by dividing 'Your financial allocation for Problem 1' by 'Total Allocation' and multiplying this with the assigned decage (similar to percentage but for 10) for Problem 1
    file_responses['Financial Token Weight for Problem 1'] = file_responses['''How much was your latest Tax payment (in U$D) ?
    
    Please try to be as accurate as possible:
    Eg.: If your last tax amount was INR 25,785/-; then convert it in U$D and enter only the amount as: 310.
    
    If you have never paid tax, consider putting in a realistic donation amount which wish to contribute towards helping yourself obtain the desired relief.'''
    ] * file_responses['''Your financial allocation for Problem 1:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a specific solution for your 1st problem.'''] / file_responses['Total Allocation']
    
    
    file_responses['Financial Token Weight for Problem 2'] = file_responses['''How much was your latest Tax payment (in U$D) ?
    
    Please try to be as accurate as possible:
    Eg.: If your last tax amount was INR 25,785/-; then convert it in U$D and enter only the amount as: 310.
    
    If you have never paid tax, consider putting in a realistic donation amount which wish to contribute towards helping yourself obtain the desired relief.'''
    ] * file_responses['''Your financial allocation for Problem 2:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 2nd problem.'''] / file_responses['Total Allocation']
    
    
    file_responses['Financial Token Weight for Problem 3'] = file_responses['''How much was your latest Tax payment (in U$D) ?
    
    Please try to be as accurate as possible:
    Eg.: If your last tax amount was INR 25,785/-; then convert it in U$D and enter only the amount as: 310.
    
    If you have never paid tax, consider putting in a realistic donation amount which wish to contribute towards helping yourself obtain the desired relief.'''
    ] * file_responses['''Your financial allocation for Problem 3:
    Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 3rd problem.'''] / file_responses['Total Allocation']

    return file_responses
    








def nlp_pipeline(original_df):
    processed_df = data_pre_processing(original_df)
    #original_df['Sum'] = original_df['a'] + original_df['b']
    return processed_df
    


def process_excel(file):
    try:
        # Ensure the file path is correct
        file_path = file.name if hasattr(file, 'name') else file
        # Read the Excel file
        df = pd.read_excel(file_path)
        
        # Perform any processing on the DataFrame here
        # Example: adding a new column with the sum of two other columns
        result_df = nlp_pipeline(df)
        
        
        return result_df  # Return the first few rows as an example
        
    except Exception as e:
        return str(e)  # Return the error message

# Define the Gradio interface
interface = gr.Interface(
    fn=process_excel,  # The function to process the uploaded file
    inputs=gr.File(type="filepath", label="Upload Excel File"),  # File upload input
    outputs="dataframe",  # Display the output as a DataFrame
    title="Excel File Uploader",
    description="Upload an Excel file to see the first few rows."
)

# Launch the interface
if __name__ == "__main__":
    interface.launch()











# #!/usr/bin/env python
# # coding: utf-8

# import pandas as pd
# import string
# import nltk
# import seaborn as sns
# import matplotlib.pyplot as plt
# from nltk.corpus import stopwords
# from nltk.tokenize import word_tokenize
# from nltk.sentiment import SentimentIntensityAnalyzer
# from sklearn.feature_extraction.text import TfidfVectorizer
# from sklearn.cluster import KMeans
# from transformers import T5ForConditionalGeneration, T5Tokenizer
# from datasets import Dataset

# # Load the data
# file_responses = pd.read_excel("#TaxDirection (Responses).xlsx")

# # Process financial allocations
# def process_allocations(df, col_name):
#     return pd.to_numeric(df[col_name], errors='coerce').fillna(0)

# columns_to_process = [
#     '''Your financial allocation for Problem 1:
# Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a specific solution for your 1st problem.''',
#     '''Your financial allocation for Problem 2:
# Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 2nd problem.''',
#     '''Your financial allocation for Problem 3:
# Mention the percentage of your Tax Amount which you wish the Government would allocate through their annual budget, to implement a solution specifically to your 3rd problem.'''
# ]

# for col in columns_to_process:
#     file_responses[col] = process_allocations(file_responses, col)

# file_responses['How much was your latest Tax payment (in U$D)?'] = pd.to_numeric(
#     file_responses['How much was your latest Tax payment (in U$D)?'], errors='coerce').fillna(0)

# # Compute total allocation and financial weights
# file_responses['Total Allocation'] = file_responses[columns_to_process].apply(lambda x: x.clip(lower=10)).sum(axis=1)

# for i in range(1, 4):
#     file_responses[f'Financial Token Weight for Problem {i}'] = (
#         file_responses['How much was your latest Tax payment (in U$D)?'] * 
#         file_responses[columns_to_process[i - 1]] / 
#         file_responses['Total Allocation']
#     )

# # Create initial datasets
# initial_datasets = []
# for i in range(1, 4):
#     initial_datasets.append(
#         file_responses[[f'''Describe Problem {i}:
# Enter the context of the problem.
# What are the difficulties you are facing personally or as a part of an organization?
# You may briefly propose a solution idea as well.''',
#             f'''Problem {i}: Geographical Location :
# Where is the location you are facing this problem?
# You may mention the nearby geographical area of the proposed solution as:
# City/Town, State/Province, Country.''',
#             f'Financial Token Weight for Problem {i}']]
#     )

# # Rename columns
# for idx, df in enumerate(initial_datasets):
#     initial_datasets[idx] = df.rename(columns={
#         df.columns[0]: 'Problem_Description',
#         df.columns[1]: 'Geographical_Location',
#         df.columns[2]: 'Financial_Weight'
#     })

# # Merge datasets
# merged_dataset = pd.concat(initial_datasets, ignore_index=True)

# # Preprocess text
# nltk.download('stopwords')
# nltk.download('punkt')
# nltk.download('omw-1.4')

# def preprocess_text(text):
#     translator = str.maketrans("", "", string.punctuation)
#     text = text.translate(translator)
#     tokens = word_tokenize(text)
#     stop_words = set(stopwords.words('english'))
#     tokens = [word for word in tokens if word.lower() not in stop_words]
#     return ' '.join(tokens)

# merged_dataset['Problem_Description'] = merged_dataset['Problem_Description'].astype(str).apply(preprocess_text)
# merged_dataset['Problem_Description'] = merged_dataset['Problem_Description'].str.replace(r'\d+', '', regex=True)
# merged_dataset['Geographical_Location'] = merged_dataset['Geographical_Location'].str.replace(r'\d+', '', regex=True)
# merged_dataset['Problem_Description'] = merged_dataset['Problem_Description'].replace(r'http\S+', '', regex=True).replace(r'www\S+', '', regex=True)
# merged_dataset['Geographical_Location'] = merged_dataset['Geographical_Location'].replace(r'http\S+', '', regex=True).replace(r'www\S+', '', regex=True)

# # Lemmatize text
# lemmatizer = nltk.WordNetLemmatizer()
# merged_dataset['Problem_Description'] = merged_dataset['Problem_Description'].apply(lambda x: ' '.join([lemmatizer.lemmatize(word) for word in x.split()]))

# # Clustering
# corpus = merged_dataset['Problem_Description'].tolist()
# tfidf_vectorizer = TfidfVectorizer(max_features=77000)
# tfidf_matrix = tfidf_vectorizer.fit_transform(corpus)

# problem_cluster_count = 77
# kmeans = KMeans(n_clusters=problem_cluster_count)
# kmeans.fit(tfidf_matrix)

# terms = tfidf_vectorizer.get_feature_names_out()
# ordered_centroids = kmeans.cluster_centers_.argsort()[:, ::-1]

# cluster_representations = {}
# for i in range(kmeans.n_clusters):
#     cluster_representations[i] = [terms[ind] for ind in ordered_centroids[i, :17]]

# merged_dataset['Problem_Category_Numeric'] = kmeans.labels_
# merged_dataset['Problem_Category_Words'] = [cluster_representations[label] for label in kmeans.labels_]

# # Clustering geographical locations
# geographical_data = merged_dataset['Geographical_Location'].tolist()
# tfidf_vectorizer_geography = TfidfVectorizer(max_features=3000)
# tfidf_matrix_geography = tfidf_vectorizer_geography.fit_transform(geographical_data)

# location_cluster_count = 33
# kmeans_locations = KMeans(n_clusters=location_cluster_count)
# kmeans_locations.fit(tfidf_matrix_geography)

# terms_geography = tfidf_vectorizer_geography.get_feature_names_out()
# ordered_centroids_geography = kmeans_locations.cluster_centers_.argsort()[:, ::-1]

# cluster_representations_geography = {}
# for i in range(kmeans_locations.n_clusters):
#     cluster_representations_geography[i] = [terms_geography[ind] for ind in ordered_centroids_geography[i, :5]]

# merged_dataset['Location_Category_Numeric'] = kmeans_locations.labels_
# merged_dataset['Location_Category_Words'] = [cluster_representations_geography[label] for label in kmeans_locations.labels_]

# # Create 2D matrices for problem descriptions and financial weights
# matrix2Dfinances = [[[] for _ in range(location_cluster_count)] for _ in range(problem_cluster_count)]
# matrix2Dproblems = [[[] for _ in range(location_cluster_count)] for _ in range(problem_cluster_count)]

# for index, row in merged_dataset.iterrows():
#     location_index = row['Location_Category_Numeric']
#     problem_index = row['Problem_Category_Numeric']
#     problem_description = row['Problem_Description']
#     financial_wt = row['Financial_Weight']
    
#     matrix2Dproblems[problem_index][location_index].append(problem_description)
#     matrix2Dfinances[problem_index][location_index].append(financial_wt)

# # Aggregating financial weights
# aggregated_Financial_wts = {}
# un_aggregated_Financial_wts = {}

# for Financ_wt_index, Financ_wt_row in enumerate(matrix2Dfinances):
#     aggregated_Financial_wts[Financ_wt_index] = {}
#     un_aggregated_Financial_wts[Financ_wt_index] = {}
    
#     for location_index, cell_finances in enumerate(Financ_wt_row):
#         cell_sum = sum(cell_finances)
#         aggregated_Financial_wts[Financ_wt_index][location_index] = cell_sum
#         un_aggregated_Financial_wts[Financ_wt_index][location_index] = cell_finances

# matrix2Dfinances_df = pd.DataFrame(aggregated_Financial_wts)
# matrix2Dfinances_df.to_excel('matrix2Dfinances_HeatMap.xlsx', index=True)

# unagregated_finances_df = pd.DataFrame(un_aggregated_Financial_wts)
# unagregated_finances_df.to_excel('UNaggregated Financial Weights.xlsx', index=True)

# # Create heatmaps
# plt.figure(figsize=(15, 7))
# sns.heatmap(matrix2Dfinances_df, annot=False, cmap='RdYlGn')
# plt.title('Project Financial Weights')
# plt.ylabel('Location Clusters')
# plt.xlabel('Problem Clusters')
# plt.savefig('Project Financial Weights_HeatMap_GreenHigh.png')
# plt.show()

# plt.figure(figsize=(14, 6))
# sns.heatmap(matrix2Dfinances_df, annot=False, cmap='RdYlGn_r')
# plt.title('Project Financial Weights')
# plt.ylabel('Location Clusters')
# plt.xlabel('Problem Clusters')
# plt.savefig('Project Financial Weights_HeatMap_RedHigh.png')
# plt.show()

# # Summarizing problems using T5
# model = T5ForConditionalGeneration.from_pretrained('t5-small')
# tokenizer = T5Tokenizer.from_pretrained('t5-small')

# def t5_summarize(text):
#     input_text = "summarize: " + text
#     inputs = tokenizer.encode(input_text, return_tensors="pt", max_length=512, truncation=True)
#     summary_ids = model.generate(inputs, max_length=150, min_length=40, length_penalty=2.0, num_beams=4, early_stopping=True)
#     return tokenizer.decode(summary_ids[0], skip_special_tokens=True)

# summarized_problems = [[t5_summarize(" ".join(cell)) for cell in row] for row in matrix2Dproblems]

# # Save summarized problems
# with open('summarized_problems.txt', 'w') as file:
#     for problem_row in summarized_problems:
#         file.write("\t".join(problem_row) + "\n")