import torch
import gradio as gr
import json
from transformers import pipeline, AutoTokenizer
print("Loading Summarization model: t5-base...")
text_summary = pipeline(
"summarization",
model="t5-base",
torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32
)
print("Summarization model loaded.")
def summarize_text(input_text):
output = text_summary(input_text, min_length=30, max_length=120)
return output[0]['summary_text']
print("Loading Translation model: facebook/nllb-200-distilled-600M...")
text_translator = pipeline(
"translation",
model="facebook/nllb-200-distilled-600M",
torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32
)
print("Translation model loaded.")
with open('language.json', 'r', encoding='utf-8') as file:
language_data = json.load(file)
def get_FLORES_code_from_language(language_name):
for entry in language_data:
if entry['Language'].lower() == language_name.lower():
return entry['FLORES-200 code']
return None
def translate_text(text, destination_language_name):
dest_code = get_FLORES_code_from_language(destination_language_name)
if not dest_code:
return f"Error: FLORES-200 code not found for '{destination_language_name}'. Please select a language from the list."
print(f"Translating from English (eng_Latn) to {destination_language_name} ({dest_code})...")
translation = text_translator(
text,
src_lang="eng_Latn",
tgt_lang=dest_code
)
return translation[0]["translation_text"]
print("Loading Question-Answering model: distilbert-base-uncased-distilled-squad...")
Youtube = pipeline(
"question-answering",
model="distilbert-base-uncased-distilled-squad",
torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32
)
print("Question-Answering model loaded.")
def read_file_content(file_obj):
try:
with open(file_obj.name, 'r', encoding='utf-8') as file:
return file.read()
except UnicodeDecodeError:
try:
with open(file_obj.name, 'r', encoding='latin-1') as file:
return file.read()
except Exception as e:
return f"An error occurred: {e}"
except Exception as e:
return f"An error occurred: {e}"
def get_answer(file, question):
context = read_file_content(file)
if "An error occurred" in context:
return context
answer = Youtube(question=question, context=context)
if answer['score'] < 0.2:
return "I am not confident I can answer this question based on the provided text."
return answer["answer"]
def landing_page():
return (
""
"
Unified NLP Toolkit ✨
"
"
"
"Summarize, translate, or ask questions on documents using a modern, friendly interface.
"
"Select a tool below to get started."
"
"
"
"
"
"
"Made with ❤️ by tushar"
"
"
)
with gr.Blocks(title="Unified NLP Toolkit ✨", theme=gr.themes.Soft(primary_hue="blue", neutral_hue="slate")) as demo:
gr.HTML(landing_page())
with gr.Tab("Text Summarizer", elem_id="summarizer"):
gr.Markdown("### Text Summarizer")
gr.Markdown("> Enter your text below and click **Summarize** to generate a concise summary.")
inp = gr.Textbox(
label="Input",
lines=10,
placeholder="Paste your text here..."
)
btn = gr.Button(
"Summarize",
elem_id="summarize-btn",
variant="secondary"
)
out = gr.Textbox(label="Summary", lines=4)
btn.click(fn=summarize_text, inputs=inp, outputs=out)
with gr.Tab("Multilanguage Translator", elem_id="translator"):
gr.Markdown("### Multilanguage Translator")
gr.Markdown("> Enter English text, select an Indian language, and click **Translate**.")
inp3 = gr.Textbox(
label="English Text",
lines=10
)
lang_options = [entry['Language'] for entry in language_data]
lang_dropdown = gr.Dropdown(
lang_options,
label="Select Indian Language"
)
btn3 = gr.Button(
"Translate",
elem_id="translate-btn",
variant="secondary"
)
out3 = gr.Textbox(label="Translated Text", lines=4)
btn3.click(fn=translate_text, inputs=[inp3, lang_dropdown], outputs=out3)
with gr.Tab("Document QnA", elem_id="qna"):
gr.Markdown("### Document Q&A")
gr.Markdown("> Upload a `.txt` file and ask a question about its content.")
file_inp = gr.File(label="Upload Text File")
q_inp = gr.Textbox(label="Your Question", lines=2)
btn4 = gr.Button(
"Get Answer",
elem_id="qna-btn",
variant="secondary"
)
out4 = gr.Textbox(label="Answer", lines=2)
btn4.click(fn=get_answer, inputs=[file_inp, q_inp], outputs=out4)
gr.HTML(footer())
demo.launch()