txtrct

app.py

@@ -2,12 +2,8 @@ import streamlit as st
 import torch
 import torch.nn.functional as F
 from torch import Tensor
-from transformers import AutoTokenizer, AutoModel
-import tempfile
+
 import textract
-import docx2txt
-import pdfplumber
-import io
 import os
 
 def last_token_pool(last_hidden_states: Tensor,
@@ -46,23 +42,9 @@ click = st.button("Search")
 
 
 
-def extract_text(doc):
-    if doc.type == 'text/plain':
-        return doc.read().decode('utf-8')
-
-    if doc.name.endswith(".pdf"):
-        docPath = save_upload(doc)
-        
-
-        with pdfplumber.open(docPath) as pdf:
-            pages = [page.extract_text() for page in pdf.pages]
 
-            return "\n".join(pages)
-    
-    
-    if doc.name.endswith('.docx'):
-        raw_text = doc.read()
-        return docx2txt.process(raw_text)
+def extract_text(doc):
+    return textract.process(doc).decode('utf-8')
 
     return None
 

requirements.txt

@@ -1,5 +1,3 @@
 torch
 transformers
-textract
-docx2txt
-pdfplumber
+textract