abstractive_summarization.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

# Function to summarize using the fine-tuned BART model
def summarize_with_bart_ft(input_text):
    pipe_bart_ft = pipeline("summarization", model="EE21/BART-ToSSimplify")
    summary = pipe_bart_ft(input_text, max_length=300, min_length=100, num_beams=1, early_stopping=False, length_penalty=1)
    return summary[0]['summary_text']

# Function to summarize using BART-large-cnn
def summarize_with_bart_cnn(input_text):
    pipe = pipeline("summarization", model="facebook/bart-large-cnn")
    summary = pipe(input_text, max_length=300, min_length=100, num_beams=1, early_stopping=False, length_penalty=1)
    return summary[0]['summary_text']

# Function to summarize using led-base-book-summary
def summarize_with_led(input_text):
    pipe_led = pipeline("summarization", model="pszemraj/led-base-book-summary")
    summary = pipe_led(input_text, max_length=300, min_length=100, num_beams=1, early_stopping=False, length_penalty=1)
    return summary[0]['summary_text']

# Function to summarize using long-t5-tglobal-base-sci-simplify
def summarize_with_t5(input_text):
    pipe_t5 = pipeline("summarization", model="pszemraj/long-t5-tglobal-base-sci-simplify")
    summary = pipe_t5(input_text, max_length=300, min_length=100, num_beams=1, early_stopping=False, length_penalty=1)
    return summary[0]['summary_text']