Spaces:
Runtime error
Runtime error
Commit
β’
6ef620e
1
Parent(s):
f095f5c
Delete app.py
Browse files
app.py
DELETED
@@ -1,77 +0,0 @@
|
|
1 |
-
# https://huggingface.co/spaces/itsmariamaraki/AAI-Assessment3
|
2 |
-
|
3 |
-
# Here are the imports
|
4 |
-
|
5 |
-
import gradio as gr
|
6 |
-
import PyPDF2
|
7 |
-
from PyPDF2 import PdfReader
|
8 |
-
from pdfminer.high_level import extract_pages, extract_text
|
9 |
-
from transformers import pipeline, AutoProcessor, AutoModel, AutoTokenizer
|
10 |
-
import torch
|
11 |
-
import soundfile as sf
|
12 |
-
from IPython.display import Audio
|
13 |
-
from datasets import load_dataset
|
14 |
-
from io import BytesIO
|
15 |
-
import os
|
16 |
-
|
17 |
-
# Here is the code
|
18 |
-
|
19 |
-
def abstract(pdf_file):
|
20 |
-
pdf_bytes = BytesIO(pdf_file)
|
21 |
-
pdf_reader = PyPDF2.PdfReader(pdf_bytes)
|
22 |
-
|
23 |
-
abstract = ''
|
24 |
-
|
25 |
-
for page_number in range(len(pdf_reader.pages)):
|
26 |
-
text = pdf_reader.pages[page_number].extract_text()
|
27 |
-
|
28 |
-
if 'abstract' in text.lower(): #in order to read only the abstract, i set as a start the abstract point & as an end the introduction point
|
29 |
-
start_index = text.lower().find('abstract')
|
30 |
-
end_index = text.lower().find('introduction')
|
31 |
-
abstract = text[start_index:end_index]
|
32 |
-
break
|
33 |
-
|
34 |
-
return abstract
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
summarization = pipeline('summarization', model = 'pszemraj/long-t5-tglobal-base-16384-book-summary') #best summarization model i tested regarding this assessment
|
39 |
-
audiospeech = pipeline('text-to-speech', model = 'suno/bark-small') #the voice is a bit distorted but gives a good output & takes less time
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
def summarization_n_audiospeech(pdf_file):
|
44 |
-
abstract_text = abstract(pdf_file)
|
45 |
-
|
46 |
-
summary = summarization(abstract_text, max_length=50, min_length=10)[0]['summary_text'] #didn't know exactly what would give one sentence, so i checked multiple times the min & max lengths regarding the 11th article. for a dif article, those parameters would probably have to be different as well
|
47 |
-
|
48 |
-
fin_summary = summary.split('.')[0] + '.' #extract and print only the first sentence of the summary
|
49 |
-
|
50 |
-
#converting the summarization into an audio output
|
51 |
-
tts_output = audiospeech(fin_summary)
|
52 |
-
audio_data = tts_output['audio'][0]
|
53 |
-
|
54 |
-
with BytesIO() as buffer:
|
55 |
-
sf.write(buffer, audio_data, 16000, format = 'wav')
|
56 |
-
audio_bytes = buffer.getvalue()
|
57 |
-
|
58 |
-
return fin_summary, audio_bytes
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
iface = gr.Interface(
|
63 |
-
fn = summarization_n_audiospeech,
|
64 |
-
inputs = gr.File(label='upload PDF', type='binary'), #if i didn't set a type, the gradio output was an error - searched it online for the solution
|
65 |
-
outputs = [
|
66 |
-
gr.Textbox(label='Summarization of the Abstract:'),
|
67 |
-
gr.Audio(label="Audio Speech of the Abstract's Summary:")
|
68 |
-
],
|
69 |
-
title = "PDF's Abstract Summarization & Audio Speech Processor",
|
70 |
-
description = "App that generates a one-line summary of the abstract & a speech audio of this summarization -- requirements: app only accepts PDFs which include an ABSTRACT section",
|
71 |
-
examples = [os.path.join(os.path.dirname(__file__), 'Hidden_Technical_Debt.pdf'),
|
72 |
-
os.path.join(os.path.dirname(__file__), 'Semiconductors.pdf'),
|
73 |
-
os.path.join(os.path.dirname(__file__), 'Efficient_Estimation_of_Word_Representations.pdf')
|
74 |
-
]
|
75 |
-
)
|
76 |
-
|
77 |
-
iface.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|