app.py

import gradio as gr
import re
from transformers import pipeline

title = "Fold: Contextual Tag Recommendation System"
description = "powered by bart-large-mnli, made by @abhisheky127"

classifier = pipeline("zero-shot-classification",
                      model="facebook/bart-large-mnli")

#define a function to process your input and output
def zero_shot(doc, candidates):
    given_labels = candidates.split(",")
    given_labels = list(map(str.strip, given_labels))
    doc = preprocess(doc)
    print(doc)
    dictionary = classifier(doc, given_labels)
    labels = dictionary['labels']
    scores = dictionary['scores']
    return dict(zip(labels, scores))

#define a function to preprocess transaction query
# def preprocess(transaction):
#     pattern = r'([A-Za-z0-9\s]+)(?:/| |$)'
#     match = re.search(pattern, transaction)
#     if match:
#         return match.group(1).strip()
#     return None

# def preprocess(transaction):
#     remove_words = ["pos", "mps", "bil", "onl"]

#     # Convert to lowercase
#     transaction = transaction.lower()

#     # Remove unwanted words
#     for word in remove_words:
#         transaction = transaction.replace(word, "")

#     # Remove special characters and digits
#     transaction = re.sub(r"[^a-z\s]+", "", transaction)

#     # Remove extra spaces
#     transaction = re.sub(r"\s+", " ", transaction).strip()
#     return transaction

def preprocess(text):
    # Remove digits
    cleaned_text = re.sub(r'\d', '', text)
    
    # Remove special characters except spaces and letters
    cleaned_text = re.sub(r'[^a-zA-Z\s]', ' ', cleaned_text)
    
    # Remove extra spaces
    cleaned_text = re.sub(r'\s+', ' ', cleaned_text).strip()
    
    # Convert to uppercase
    cleaned_text = cleaned_text.upper()
    
    # Remove unwanted words
    words_to_remove = ["MPS", "POS", "BIL", "ONL", "BANGALORE", "PVT", "LTD", "INDIA", "LT"]
    cleaned_text = " ".join([word for word in cleaned_text.split() if word not in words_to_remove])
    
    # Convert to lowercase
    cleaned_text = cleaned_text.lower()
    
    return cleaned_text


#create input and output objects
#input object1
input1 = gr.Textbox(label="Text")

#input object 2
input2 = gr.Textbox(label="Labels")

#output object
output = gr.Label(label="Output")

#example object
transactions_and_tags = [
    ["MPS/TRUFFLES/202303261700/034587/Bangalore", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["MPS/TACO BELL/202304012247/108300/BANGALORE", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS XXXXXXXXXXXX0001APOLLO PHARMACY", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["BIL/ONL/000471093694/1MG Techno/X7ZRUSVLURFQZO", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS XXXXXXXXXXXX1111 DECATHLON SPORTS", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS XXXXXXXXXXXX1111 IKEA INDIA PVT L", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS XXXXXXXXXXXX1111 WWW AMAZON IN", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["ME DC SI XXXXXXXXXXXX1111 SPOTIFY SI", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS/NETFLIX/1140920002/100623/17:25", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["POS XXXXXXXXXXXX1110 MAKEMYTRIP INDIA", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"],
    ["BIL/ONL/000691178015/IRCTC Serv/XZZBX91LTCY1AZ", "Medical, Food, Shopping, Subscription, Travel, Miscellaneous"]
]

#create interface
gui = gr.Interface(title=title,
                   description=description,
                   fn=zero_shot,
                   inputs=[input1, input2],
                   outputs=[output],
                   examples=transactions_and_tags)

#display the interface
gui.launch(debug=True)