Create app.py

app.py

@@ -0,0 +1,188 @@
+from huggingface_hub import hf_hub_download
+import joblib
+
+repo_id = "DevBhojani/Classification-SamsumDataset"
+model_filename = "random_forest_classifier_model.joblib"
+
+model_path = hf_hub_download(repo_id=repo_id, filename=model_filename)
+loaded_classifier_model = joblib.load(model_path)
+
+import joblib
+from sklearn.feature_extraction.text import TfidfVectorizer
+
+repo_id = "DevBhojani/Classification-SamsumDataset"
+model_filename = "random_forest_classifier_model.joblib"
+vectorizer_filename = "tfidf_vectorizer.joblib"
+
+model_path = hf_hub_download(repo_id=repo_id, filename=model_filename)
+vectorizer_path = hf_hub_download(repo_id=repo_id, filename=vectorizer_filename)
+
+loaded_classifier_model = joblib.load(model_path)
+loaded_tfidf_vectorizer = joblib.load(vectorizer_path)
+
+import gradio as gr
+from transformers import pipeline, AutoTokenizer
+import re
+import contractions
+# Assuming loaded_classifier_model and loaded_tfidf_vectorizer are already loaded from the previous cell
+
+def remove_html_tags(text):
+    pattern = r'[^a-zA-Z0-9\s]'
+    text = re.sub(pattern, '', str(text))
+    return text
+
+def remove_url(text):
+    pattern = re.compile(r'https?://\S+|www\.\S+')
+    return pattern.sub(r'', str(text))
+
+def remove_emojis(text):
+    emoji_pattern = re.compile(
+        "["
+        u"\U0001F600-\U0001F64F"  # emoticons
+        u"\U0001F300-\U0001F5FF"  # symbols & pictographs
+        u"\U0001F680-\U0001F6FF"  # transport & map symbols
+        u"\U0001F1E0-\U0001F1FF"  # flags
+        u"\U00002700-\U000027BF"  # miscellaneous symbols
+        u"\U0001F900-\U0001F9FF"  # supplemental symbols
+        u"\U00002600-\U000026FF"  # weather & other symbols
+        u"\U0001FA70-\U0001FAFF"  # extended symbols
+        "]+",
+        flags=re.UNICODE
+    )
+    return emoji_pattern.sub(r'', str(text))
+
+def expand_contractions(text):
+    return contractions.fix(text)
+
+def remove_special_and_numbers(text):
+    return re.sub(r'[^a-zA-Z\s]', '', str(text))
+
+def clean_text(text):
+    text = remove_url(text)
+    text = remove_emojis(text)
+    text = expand_contractions(text)
+    text = text.lower()
+    return text
+
+summarizer = pipeline("summarization", model="luisotorres/bart-finetuned-samsum")
+# summarizer2 = pipeline("summarization", model="knkarthick/MEETING_SUMMARY")
+tokenizer = AutoTokenizer.from_pretrained("luisotorres/bart-finetuned-samsum")
+
+def split_into_chunks(conversation, n=15):
+    lines = conversation.strip().split('\n')
+    chunk_size = max(1, len(lines) // n)
+    return ['\n'.join(lines[i:i+chunk_size]) for i in range(0, len(lines), chunk_size)]
+
+def truncate_chunk(text, max_tokens=1024):
+    tokens = tokenizer.encode(text, truncation=True, max_length=max_tokens)
+    return tokenizer.decode(tokens, skip_special_tokens=True)
+
+def summarize_chunks(chunks, model):
+    summaries = []
+    for chunk in chunks:
+        chunk = chunk.strip()
+        if not chunk:
+            continue
+        try:
+            truncated_chunk = truncate_chunk(chunk)
+            summary = model(truncated_chunk, max_length=1024, min_length=20, do_sample=False)[0]['summary_text']
+            summaries.append(summary)
+        except Exception as e:
+            print(f"Error summarizing chunk: {e}")
+    return summaries
+
+def combine_summaries(summaries):
+    return ' '.join(summaries)
+
+def summarize_dialogue(conversation, model):
+    chunks = split_into_chunks(conversation, n=1)
+    summaries = summarize_chunks(chunks, model)
+    final_summary = combine_summaries(summaries)
+    return final_summary
+
+def analyze_meeting_transcript(user_input):
+    if not user_input.strip():
+        return "Please enter some text to summarize.", ""
+
+    cleaned_input = clean_text(user_input)
+    summary1 = summarize_dialogue(cleaned_input, summarizer)
+
+    # Use the loaded vectorizer to transform the input
+    cleaned_input_vectorized = loaded_tfidf_vectorizer.transform([cleaned_input])
+
+    intent_classification = loaded_classifier_model.predict(cleaned_input_vectorized)[0]
+    # print(intent_classification)
+    # print(cleaned_input_vectorized)
+    # intent_classification = "Transactional Inquiry & Information Exchange"
+
+    # Format the intent classification output
+    formatted_intent = intent_classification.replace("__label__", "").replace("_", " ")
+
+
+    return summary1, formatted_intent
+
+interface = gr.Interface(
+    fn=analyze_meeting_transcript,
+    inputs=gr.Textbox(label="Enter dialogue here", lines=12, placeholder="Paste your meeting transcript..."),
+    outputs=[
+        gr.Textbox(label="Summary (Luis Torres BART)"),
+        # gr.Textbox(label="Summary 2 (KN Karthick MEETING_SUMMARY)"),
+        gr.Textbox(label="Intent Classification") # Removed "Placeholder"
+    ],
+    title="Meeting Transcript Analyzer",
+    description="Summarizes meeting dialogues and classifies the intent.",
+    allow_flagging="never",
+    examples=[
+        [
+            '''
+Amanda: guess what!
+Chris: hey ;) ur pregnant!
+Amanda: noo ;) but close enough! I'm so proud of myself! Remember I go to these dancing classes with Michael?
+Chris: Yeah?
+Amanda: So we went yesterday and the instructor needed a partner to show the steps we had so far
+Chris: so there's only one guy teaching you? without a female partner?
+Amanda: Well, this time he was alone, BUT THAT'S NOT THE POINT! Listen!
+Chris: yeah, sorry :D tell me!
+Amanda: So he needed a partner and noone really knew the steps like perfectly
+Amanda: and obviously noone wanted to be mocked
+Amanda: so I thought, aaaah :D
+Chris: u volunteered? really? you??
+Amanda: yeah!
+Chris: whooa! that's so great! #therapy #worthit :D
+Amanda: yeah i know :D maybe one day i'll actually stop being so shy
+Chris: that's definitely the first step! :D congrats!
+Amanda: tx ^_^
+Chris: what dance was it?
+Amanda: English waltz
+Chris: isn't it, like, SO difficult?
+Amanda: yeah it is! but everyone said I looked like a pro :D
+Chris: Well done!!
+'''
+        ],
+        ["I have some exciting news to share!"],
+        [
+            '''
+Beryl: Hello guys! How are you doing? We've lost contact for a few months now. Hope you are well.
+Anton: A happy hello to you Beryl! Great to hear from you. We are fine, thanks. And yourself?
+Beryl: I'm very well indeed. Thank you. Any changes in your setup?
+Anton: Not really. SOS. Same Old Soup ;) But we are happy for that.
+Beryl: Are you still running your lovely airbnb?
+Anton: Oh yes, we are. We had a few months off during summer, our summer, but now bookings start flowing in. Well... Are you planning to visit us? You two are always welcome!
+Beryl: You caught me here. I'm vaguely considering going down to Onrus again, most likely in January. What does it look like with vacancies then?
+Anton: Perfect! Just give me your dates and I'll keep it booked for you.
+Beryl: Would you prefer me to do it via airbnb website or just like this directly with you?
+Anton: I think it'll be more advantageous for both of us to do it directly. Do you know exactly when you'll be coming?
+Beryl: Not so much. Can I get back to you in 2, 3 days' time?
+Anton: ASAP really. As I say we've been receiving bookings daily now.
+Beryl: Well, no big deal. I'll be staying in Cape Town for a longer time and am quite flexible in my dates.
+Anton: Will you be coming with Tino, if I may ask?
+Beryl: No. I am single again. Hurray! So pls make it single occupancy any week in January, Anton.
+Anton: Great! 4th till 12th?
+Beryl: Very good. I'll call you beforehand from Cape Town. Greetings to you both!
+Anton: Take care!'''
+        ],
+    ]
+)
+
+if __name__ == "__main__":
+    interface.launch(debug=True, share=True)