Update app.py

app.py

@@ -1,46 +1,37 @@
 from flask import Flask, request, jsonify
-from sentence_transformers import SentenceTransformer, util
-import torch
-import re
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
 
 app = Flask(__name__)
 
-# Load model SBERT tiếng Việt
-retriever = SentenceTransformer("keepitreal/vietnamese-sbert")
+# --- Load mô hình tóm tắt tiếng Việt ---
+model_name = "NlpHUST/t5-small-vi-summarization"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-def extract_summary(context, topic, k=5, threshold=0.15):
-    raw_sentences = [s.strip() for s in re.split(r'(?<=[.!?])\s+', context) if s.strip()]
-    if len(raw_sentences) == 0:
-        return ""
-    context_emb = retriever.encode(raw_sentences, convert_to_tensor=True)
-    query_emb = retriever.encode(topic, convert_to_tensor=True)
-    cosine_scores = util.pytorch_cos_sim(query_emb, context_emb)[0]
-
-    k_eff = min(k, len(raw_sentences))
-    topk = torch.topk(cosine_scores, k=k_eff)
-    top_indices = topk.indices.tolist()
-    top_scores = topk.values.tolist()
-
-    filtered = [(i, s) for i, s in zip(top_indices, top_scores) if s >= threshold]
-    if len(filtered) == 0:
-        filtered = [(top_indices[0], top_scores[0])]
-    filtered.sort(key=lambda x: x[0])
-    selected_text = " ".join([raw_sentences[i] for i, _ in filtered])
-    return selected_text
+summarizer = pipeline(
+    "summarization",
+    model=model,
+    tokenizer=tokenizer
+)
 
 @app.route("/")
 def home():
-    return jsonify({"message": "✅ AI Extractive Text Summarization API (Vietnamese) running."})
+    return jsonify({
+        "message": "✅ AI Text Summarization API is running (Vietnamese, NlpHUST/t5-small-vi).",
+        "usage": "POST /summarize với JSON: { 'text': '...', 'topic': '...' }"
+    })
 
 @app.route("/summarize", methods=["POST"])
 def summarize():
     data = request.get_json(force=True)
-    if not data or "text" not in data or "topic" not in data:
-        return jsonify({"error": "Thiếu trường 'text' hoặc 'topic'"}), 400
+    if not data or "text" not in data:
+        return jsonify({"error": "Thiếu trường 'text'"}), 400
+    
     text = data["text"]
-    topic = data["topic"]
+    # topic không cần thiết, mô hình này tự tóm tắt
     try:
-        summary = extract_summary(text, topic)
+        result = summarizer(text, max_length=150, min_length=40, do_sample=False)
+        summary = result[0]['summary_text']
         return jsonify({"summary": summary})
     except Exception as e:
         return jsonify({"error": str(e)}), 500