Update app.py

app.py

@@ -1,19 +1,16 @@
 
 import gradio as gr
 import json
-import spacy
 import re
 import string
 import pandas as pd
 import os
-os.system('python -m spacy download en_core_web_sm')
 import requests
 from textwrap import wrap
 import uuid
 import gspread
+import ast
 
-nlp = spacy.load("en_core_web_sm")
-nlp.add_pipe('sentencizer')
 
 
 def download_and_save_file(URL, audio_dir):
@@ -48,32 +45,52 @@ description = '''Choose a sentence that describes audio the best if there's no s
 audio_dir = 'AUDIO'
 os.makedirs(audio_dir, exist_ok=True)
 
-
-
-gc = gspread.service_account(filename='credentials.json')
-sh = gc.open('Annotated CC Audio')
-worksheet = sh.sheet1
-df = pd.DataFrame(worksheet.get_all_records())
-sample_df = df[df['caption']==''].sample(1)
-
-url, audio_url, _, _, full_text, _, _ = sample_df.values[0]
-audio_path = download_and_save_file(audio_url, audio_dir)
-full_text = full_text.translate(str.maketrans('', '', string.punctuation))
-sents = ['\n'.join(wrap(re.sub(r'###audio###\d###', '', s.text), width=70) )for s in nlp(full_text).sents]
-sents.append('No audio description')
-
-def audio_demo(cap, audio, annotator, audio_url):
+def sample_df():
+
+    gc = gspread.service_account(filename='credentials.json')
+    sh = gc.open('Annotated CC Audio')
+    worksheet = sh.sheet1
+    df = pd.DataFrame(worksheet.get_all_records())
+    sample_df = df[df['caption']==''].sample(1)
+
+    audio_url, audio_meta, page_title, img_metadata, sibling_elems = sample_df[['audio_url', 'audio_meta', 'page_title', 'imgs_metadata', 'sibling_elems']].values[0]
+    audio_path = download_and_save_file(audio_url, audio_dir)
+    sibling_elems = ast.literal_eval(sibling_elems)
+    sibling_elems = [s.replace('\n', '') for s in sibling_elems]
+    sibling_elems = [s for s in sibling_elems if len(s) > 0]
+    sibling_elems = list(set(sibling_elems))
+    img_metadata = ast.literal_eval(img_metadata)
+    if len(img_metadata) > 0:
+      img_metadata = [[f'{k}: {meta[k]}' for k in meta] for meta in img_metadata]
+    audio_meta = ast.literal_eval(audio_meta).get('tags', None)
+    if audio_meta:
+      audio_meta = [f'{k}: {audio_meta[k]}' for k in audio_meta.keys() if k.lower() in ['title', 'album', 'artist', 'genre', 'date', 'language']]
+      audio_meta = '; '.join(audio_meta)
+    return audio_path, audio_url, sibling_elems, audio_meta, page_title, df, worksheet
+
+def audio_demo(siblings, page_title, audio_meta, audio, annotator, audio_url):
     annotator = annotator if annotator else str(uuid.uuid4())
-
+    siblings.extend(page_title)
+    siblings.extend(audio_meta)
+    siblings = [s for s in siblings if s!=[]]
+    cap = '\n'.join(siblings)
     df['caption'].loc[df['audio_url'] == audio_url] = cap
     df['annotator'].loc[df['audio_url'] == audio_url] = annotator
     worksheet.update([df.columns.values.tolist()] + df.values.tolist())
     return 'success!'
 
+audio_path, audio_url, sibling_elems, audio_meta, page_title, df, worksheet = sample_df()
 
 iface = gr.Interface(
     audio_demo,  
-    inputs=[gr.Radio(sents, label='audio description'), gr.Audio(audio_path, type="filepath"), gr.Textbox(label='please enter your name'), gr.Textbox(value=audio_url, visible=False)],
+    inputs=[
+        gr.CheckboxGroup(sibling_elems, label='sibling elements text'), 
+        gr.CheckboxGroup(label='page title', choices=[page_title]), 
+        gr.CheckboxGroup([audio_meta], label='audio metadata'), 
+        gr.Audio(audio_path, type="filepath"), 
+        gr.Textbox(label='please enter your name'), 
+        gr.Textbox(value=audio_url, visible=False)
+        ],
     outputs=[gr.Textbox(label="output")],
     allow_flagging="never",
     title=title,