Update app.py

app.py

@@ -1,19 +1,17 @@
 import requests
-import pandas as pd
-from transformers import AutoTokenizer, AutoModelForSequenceClassification
-import torch
-import torch.nn.functional as F
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
+import csv
 from datetime import datetime
 import gradio as gr
 
 # ==========================
-# 1️⃣ MODEL LOAD (DistilFinBERT)
+# 1️⃣ MODEL (CPU-Friendly)
 # ==========================
-MODEL_NAME = "yiyanghkust/finbert-tone-distil"  # DistilFinBERT, CPU-friendly
-
+MODEL_NAME = "cardiffnlp/twitter-roberta-base-financial-sentiment"
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
 model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)
-model.eval()
+
+nlp = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
 
 # ==========================
 # 2️⃣ REUTERS SCRAPER
@@ -43,52 +41,53 @@ def fetch_reuters_articles(section="/business/", size=10):
     except Exception as e:
         print("Parsing error:", e)
 
-    return pd.DataFrame(articles)
+    return articles
 
 # ==========================
 # 3️⃣ SENTIMENT SCORING
 # ==========================
-def finbert_score(text):
-    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
-    with torch.no_grad():
-        outputs = model(**inputs)
-        probs = F.softmax(outputs.logits, dim=1)
-    labels = ["negative", "neutral", "positive"]
-    scores = probs[0].tolist()
-    return dict(zip(labels, scores))
+def score_articles(articles):
+    scored = []
+    for art in articles:
+        text = art["headline"] + ". " + art["summary"]
+        result = nlp(text)[0]  # cardiffnlp döndürür: {'label': 'positive', 'score': 0.95}
+        art.update(result)
+        scored.append(art)
+    return scored
 
 # ==========================
-# 4️⃣ PIPELINE
+# 4️⃣ CSV SAVE
 # ==========================
-def run_pipeline():
-    df = fetch_reuters_articles(size=10)
-    sentiments = []
-    for _, row in df.iterrows():
-        text = row["headline"] + ". " + row["summary"]
-        score = finbert_score(text)
-        sentiments.append(score)
+def save_csv(scored_articles):
+    filename = f"reuters_sentiment_{datetime.today().date()}.csv"
+    keys = scored_articles[0].keys()
+    with open(filename, "w", newline="", encoding="utf-8") as f:
+        dict_writer = csv.DictWriter(f, fieldnames=keys)
+        dict_writer.writeheader()
+        dict_writer.writerows(scored_articles)
+    return filename
 
-    sentiment_df = pd.DataFrame(sentiments)
-    final_df = pd.concat([df, sentiment_df], axis=1)
-    final_df["hawkish_score"] = final_df["negative"] - final_df["positive"]
+# ==========================
+# 5️⃣ PIPELINE
+# ==========================
+def run_pipeline(dummy_input):
+    articles = fetch_reuters_articles(size=10)
+    scored = score_articles(articles)
+    csv_file = save_csv(scored)
 
-    filename = f"reuters_finbert_{datetime.today().date()}.csv"
-    final_df.to_csv(filename, index=False)
-    return final_df.head(), filename
+    # Terminal için tablo görünümü
+    display_text = "\n".join([f"{a['headline']} -> {a['label']} ({a['score']:.2f})" for a in scored])
+    return display_text, f"Saved CSV: {csv_file}"
 
 # ==========================
-# 5️⃣ GRADIO WEB APP
+# 6️⃣ GRADIO WEB APP
 # ==========================
-def gradio_interface(dummy_input):
-    table, fname = run_pipeline()
-    return table.to_string(), f"Saved CSV: {fname}"
-
 iface = gr.Interface(
-    fn=gradio_interface,
+    fn=run_pipeline,
     inputs=gr.Textbox(label="Run Scraper (any text)"),
     outputs=[gr.Textbox(label="Top Articles"), gr.Textbox(label="CSV Status")],
-    title="Reuters DistilFinBERT Scraper",
-    description="Fetch latest Reuters Business news + sentiment via DistilFinBERT"
+    title="Reuters Financial Sentiment (CardiffNLP)",
+    description="Fetch latest Reuters Business news + sentiment via CardiffNLP Roberta model (CPU-friendly, no pandas)"
 )
 
 if __name__ == "__main__":