pages/1_Classification.py

#importing the libraries needed
#for the classification
import streamlit as st
import pandas as pd
from transformers import pipeline
#for the pre-processing of uploaded data
import string 
#adding the css presentation from the styles file in utils
from utils.styles import load_css
load_css()

#calling and loading the fine tuned model from the hugging face account
@st.cache_resource
def load_classifier():
    return pipeline(
        "text-classification",
        model="SuchithTissera/text_analytics_model",
        truncation=True
    )
#adding stop word based on nlk library - look at limitations for the reason
stop_words = {'most', "i've", 't', "we'll", 'mustn', 'we', 'into', 'they', 'on', "they'll", 
              'haven', 'after', 'had', 'myself', 'she', 'too', 'won', 'over', 'any', "you've", 
              'not', 'me', "they're", "wouldn't", 'other', 'have', 'once', 'yourselves', 'doing', 
              'mightn', 'been', "he'll", "she'll", 'the', "wasn't", 'to', 'further', "hadn't", "i'll", 
              'my', 'or', 'against', "doesn't", 'can', 'you', 'down', 'of', 'ain', 'below', 'these', "they'd", 
              'but', 'here', 'was', 'and', 'both', 'so', 'your', 'needn', 'has', "she's", 'as', "should've", 
              'very', 've', 'them', "aren't", "shouldn't", 'where', 'while', 'our', "mustn't", 'yourself', 'by', 
              'now', 'be', 'do', 'y', 'should', 'about', 'for', 'her', 'then', 'from', "isn't", "he'd", 'whom', 
              'herself', 'before', "he's", "you'd", 'm', "didn't", "we've", 'no', "hasn't", "needn't", 'off', 's', 
              'some', 'are', 'all', "weren't", "i'm", 'theirs', 'wasn', 'during', 'didn', "haven't", 'own', 'at', 
              'hasn', 'his', 'shan', "it's", 'just', 'ourselves', 'between', 'its', 'again', "she'd", 're', 'that', 
              'few', 'more', 'hadn', 'is', 'nor', "you'll", 'having', 'than', "i'd", 'only', 'aren', "you're", "couldn't",
              "won't", 'each', 'if', 'am', 'hers', 'what', 'weren', 'will', 'an', 'it', "they've", "shan't", "that'll", 'll', 
              'he', 'how', "we'd", 'did', 'isn', 'in', 'ma', 'o', 'themselves', "don't", 'shouldn', 'those', 'why', 'who',
              "we're", "it'd", 'wouldn', 'don', 'until', 'up', 'out', 'couldn', 'yours', "it'll", 'because', 'there', 
              'a', 'same', 'which', 'being', 'himself', 'under', 'through', 'when', "mightn't", 'ours', 'above', 'were', 
              'd', 'does', 'such', 'doesn', 'itself', 'this', 'i', 'their', 'with', 'him'}

#constructing the stopwork function
def remove_stopwords(text):
    return " ".join([word for word in str(text).lower().split() if word not in stop_words])

#saving the loaded model for convience in a variable 
classifier = load_classifier()

#writing the text content in the page
st.markdown('<div class="main_title">📂 News Classifier </div>',unsafe_allow_html=True)
st.markdown("""<div class="feature_text">In this page the user can upload a CSV file which contains a content 
        column which includes text, then the model will use text pre-processing (anlysis step for more details) 
        and then label the data based on 5 categories,</div>""",unsafe_allow_html=True)
st.markdown("""
<div class="feature_text">
<ul>
    <li>Business</li>
    <li>Opinions</li>
    <li>Political_gossip</li>
    <li>Sports</li>
    <li>Worlds_news</li>
</ul>
</div>
""", unsafe_allow_html=True)

#dividing the page
st.divider()

#uploading the file 
st.markdown(
    '<div class="feature_text">Upload your CSV file:</div>',
    unsafe_allow_html=True
)

#saving the csv file in a variable
uploaded_file = st.file_uploader("", type=["csv"])

#loop to carry out the transformation
if uploaded_file is not None:
    try:
        uploaded_content = pd.read_csv(uploaded_file)
        #detect if a content coloumn is present
        if "content" not in uploaded_content.columns:
            st.error("The uploaded CSV must contain a column named `content`.")
        else:
            #removing empty strings
            uploaded_content["content"] = uploaded_content["content"].fillna("").astype(str)
            #saving the pre-processed datapoints as lowercase 
            uploaded_content["pre-processed"] = uploaded_content["content"].str.lower()
            #showing the loading message to user 
            with st.spinner("Classifying records..."):

                #processing the data
                #removing punctuations
                def remove_punctuation(text):
                    return text.translate(str.maketrans('', '', string.punctuation))
                uploaded_content["pre-processed"] = uploaded_content["pre-processed"].apply(remove_punctuation)
                #removing stop words 
                uploaded_content["pre-processed"] = uploaded_content["pre-processed"].apply(remove_stopwords)

                #constructing the labels
                preds = classifier(uploaded_content["pre-processed"].tolist(), batch_size=8)

            uploaded_content["class"] = [pred["label"] for pred in preds]

            # save for other pages
            st.session_state["uploaded_content"] = uploaded_content
            st.session_state["uploaded_filename"] = uploaded_file.name

            st.success("Classification complete.")
            st.dataframe(uploaded_content.head())

            #converting to a csv
            output_csv = uploaded_content.to_csv(index=False).encode("utf-8")

            #download action
            st.download_button(
                label="Download output.csv",
                data=output_csv,
                file_name="output.csv",
                mime="text/csv"
            )
    #exception handling if errors happpen
    except Exception as e:
        st.error(f"Error reading or processing file: {e}")