app.py

import time

import pandas as pd
import streamlit as st
from transformers import pipeline
from constants import tweet_generator_prompt, absa_prompt


# Initialize the model and tokenizer once, to avoid reloading them on each user interaction
@st.cache_resource
def load_model():
    start = time.time()
    classification_pipe = pipeline(
        "text-classification", model="tweetpie/toxic-content-detector", top_k=None)
    print(f"Time to load the classification model: {time.time() - start:.2f}s")
    start = time.time()
    absa_pipe = pipeline("text2text-generation", model="tweetpie/stance-aware-absa")
    print(f"Time to load the absa model: {time.time() - start:.2f}s")
    start = time.time()
    tweet_generation_pipe = pipeline("text2text-generation", model="tweetpie/stance-directed-tweet-generator")
    print(f"Time to load the tweet generation model: {time.time() - start:.2f}s")
    return classification_pipe, absa_pipe, tweet_generation_pipe


# Set up the title
st.title("Toxic Content Classifier Dashboard")

# Top-level input for model selection
model_selection = st.selectbox(
    "Select an ideology",
    options=['Left', 'Right'],
    index=0  # Default selection
)

# Layout for entities and aspects inputs
col1, col2 = st.columns(2)

with col1:
    st.header("Entities")
    pro_entities = st.text_input("Pro Entities", help="Enter pro entities separated by commas")
    anti_entities = st.text_input("Anti Entities", help="Enter anti entities separated by commas")
    neutral_entities = st.text_input("Neutral Entities", help="Enter neutral entities separated by commas")

with col2:
    st.header("Aspects")
    pro_aspects = st.text_input("Pro Aspects", help="Enter pro aspects separated by commas")
    anti_aspects = st.text_input("Anti Aspects", help="Enter anti aspects separated by commas")
    neutral_aspects = st.text_input("Neutral Aspects", help="Enter neutral aspects separated by commas")

# Generate button
generate_button = st.button("Generate and classify toxicity")

# Load the model
classifier, absa, generator = load_model()

# Process the input text and generate output
if generate_button:
    with st.spinner('Generating the tweet...'):
        # Call the model with the aspects inputs
        prompt = tweet_generator_prompt.format(
            ideology=model_selection.lower(),
            pro_entities=pro_entities,
            anti_entities=anti_entities,
            neutral_entities=neutral_entities,
            pro_aspects=pro_aspects,
            anti_aspects=anti_aspects,
            neutral_aspects=neutral_aspects
        )
        time.sleep(5)
        generated_tweet = [{"generated_text": "the agricultural sector is the single biggest recipient of migrants workers rights groups argue . nearly 90 % of those who come to the us are denied employment due to discriminatory employment laws and safety standards ."}]
        # print("Prompt: ", prompt)
        # start = time.time()
        # generated_tweet = generator(prompt, max_new_tokens=80, do_sample=True, num_return_sequences=3)
        # print(f"Time to generate the tweet: {time.time() - start:.2f}s")

        # Displaying the input and model's output
        st.write(f"Generated Tweet: {generated_tweet[0]['generated_text']}")

    with st.spinner('Generating the Stance-Aware ABSA output...'):
        # Call the model with the aspects inputs
        absa_output = absa(absa_prompt.format(generated_tweet=generated_tweet[0]['generated_text']))
        print("ABSA Output: ", absa_output)
        stances = [x.strip() for x in absa_output[0]['generated_text'].split(',')]
        stances = [{
            'Aspect': x.split(':')[0],
            'Sentiment': x.split(':')[1]
        } for x in stances]
        stances_df = pd.DataFrame(stances)
        st.write("Stance-Aware ABSA Output:")
        st.table(stances_df)

    with st.spinner('Classifying the toxicity...'):
        # Call the model with the input text
        model_output = classifier(generated_tweet[0]['generated_text'])
        output = model_output[0]

        st.write("Toxicity Classifier Output:")
        for i in range(3):
            if output[i]['label'] == 'LABEL_0':
                st.write(f"Non-Toxic Content: {output[i]['score']*100:.1f}%")
                # print(f"Non-Toxic Content: {output[i]['score']*100:.1f}%")
            elif output[i]['label'] == 'LABEL_1':
                st.write(f"Toxic Content: {output[i]['score']*100:.1f}%")
                # print(f"Toxic Content: {output[i]['score']*100:.1f}%")
            else:
                continue