Spaces:
Sleeping
Sleeping
File size: 6,693 Bytes
7761e22 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 | import re
from pathlib import Path
import gradio as gr
import numpy as np
import pdfplumber
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
from transformers import pipeline
# ---------- Models ----------
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2")
# ---------- Global state (will be stored in gr.State) ----------
# lecture_chunks, vectorizer, X_matrix will live in state
# ---------- Helpers ----------
def load_text_from_file(file_obj) -> str:
if file_obj is None:
return ""
path = Path(file_obj.name)
suffix = path.suffix.lower()
if suffix == ".pdf":
texts = []
with pdfplumber.open(file_obj) as pdf:
for page in pdf.pages:
page_text = page.extract_text() or ""
texts.append(page_text)
raw_text = "\n".join(texts)
elif suffix == ".txt":
raw_text = file_obj.read().decode("utf-8", errors="ignore")
else:
raise ValueError("Only .pdf and .txt files are supported.")
return clean_text(raw_text)
def clean_text(text: str) -> str:
text = text.replace("\r", " ")
text = re.sub(r"\n+", "\n", text)
text = re.sub(r"[ \t]+", " ", text)
return text.strip()
def chunk_text(text: str, chunk_words: int = 350, overlap_words: int = 50):
words = text.split()
chunks = []
start = 0
chunk_id = 1
while start < len(words):
end = start + chunk_words
chunk_words_list = words[start:end]
chunk_text_ = " ".join(chunk_words_list)
chunks.append(
{
"chunk_id": f"C{chunk_id}",
"text": chunk_text_,
}
)
chunk_id += 1
start = end - overlap_words
return chunks
def build_retriever(chunks):
docs = [c["text"] for c in chunks]
vectorizer = TfidfVectorizer(
max_features=10000,
ngram_range=(1, 2),
min_df=1,
)
X = vectorizer.fit_transform(docs)
return vectorizer, X
def generate_summary(text: str, max_words: int = 300) -> str:
if not text:
return "No text found in the uploaded file."
# Hugging Face summarization has a max token limit; we slice text roughly
# into smaller windows and summarize each, then summarize again.
# Keep it simple & fast.
max_chunk_chars = 2500
windows = []
start = 0
while start < len(text):
end = start + max_chunk_chars
windows.append(text[start:end])
start = end
partial_summaries = []
for w in windows[:3]: # hard cap, don’t explode runtime
s = summarizer(
w,
max_length=180,
min_length=60,
do_sample=False,
truncation=True,
)[0]["summary_text"]
partial_summaries.append(s)
combined = " ".join(partial_summaries)
final = summarizer(
combined,
max_length=220,
min_length=80,
do_sample=False,
truncation=True,
)[0]["summary_text"]
return final
def retrieve_chunks(question, chunks, vectorizer, X, top_k: int = 5):
if not chunks or vectorizer is None or X is None:
return []
q_vec = vectorizer.transform([question])
sims = cosine_similarity(q_vec, X)[0]
top_idx = np.argsort(-sims)[:top_k]
results = []
for rank, idx in enumerate(top_idx, start=1):
c = chunks[idx]
results.append(
{
"rank": rank,
"chunk_id": c["chunk_id"],
"text": c["text"],
"similarity": float(sims[idx]),
}
)
return results
def answer_question(question, chunks, vectorizer, X):
if not question.strip():
return "Please enter a question.", ""
retrieved = retrieve_chunks(question, chunks, vectorizer, X, top_k=3)
if not retrieved:
return "Please upload and process a lecture first.", ""
context_text = "\n\n".join([r["text"] for r in retrieved])
try:
ans = qa_pipeline(
{
"question": question,
"context": context_text,
}
)
answer = ans.get("answer", "").strip()
except Exception as e:
answer = f"Error from QA model: {e}"
# Build a short “sources” string
source_info = "; ".join(
[f"{r['chunk_id']} (sim={r['similarity']:.3f})" for r in retrieved]
)
return answer, source_info
# ---------- Gradio Callbacks ----------
def process_lecture(file):
"""
1. Read PDF/TXT
2. Chunk
3. Build retriever
4. Generate summary
Returns: summary, chunks, vectorizer, X
"""
if file is None:
return "Please upload a lecture file.", [], None, None
try:
text = load_text_from_file(file)
except Exception as e:
return f"Error reading file: {e}", [], None, None
if len(text) < 100:
return "File text is too short or empty after extraction.", [], None, None
chunks = chunk_text(text, chunk_words=350, overlap_words=50)
vectorizer, X = build_retriever(chunks)
summary = generate_summary(text)
return summary, chunks, vectorizer, X
def chat_fn(question, chunks, vectorizer, X):
answer, sources = answer_question(question, chunks, vectorizer, X)
if sources:
answer = f"{answer}\n\n_Sources: {sources}_"
return answer
# ---------- Gradio UI ----------
with gr.Blocks() as demo:
gr.Markdown("# 📚 Lecture Summarizer + Chatbot\nUpload a PDF/TXT lecture, get a summary, then ask questions about it.")
with gr.Row():
file_input = gr.File(label="Upload lecture (.pdf or .txt)")
process_btn = gr.Button("Process Lecture")
summary_box = gr.Textbox(
label="Lecture Summary",
lines=12,
interactive=False,
)
# State: saved across chat turns
chunks_state = gr.State([])
vectorizer_state = gr.State(None)
X_state = gr.State(None)
process_btn.click(
fn=process_lecture,
inputs=[file_input],
outputs=[summary_box, chunks_state, vectorizer_state, X_state],
)
gr.Markdown("## 💬 Chat with the Lecture")
with gr.Row():
question_box = gr.Textbox(label="Your Question")
answer_box = gr.Textbox(label="Answer", lines=6, interactive=False)
ask_btn = gr.Button("Ask")
ask_btn.click(
fn=chat_fn,
inputs=[question_box, chunks_state, vectorizer_state, X_state],
outputs=[answer_box],
)
if __name__ == "__main__":
demo.launch()
|