Enter financial texts, one per line.
"
if classifier is None:
return create_error_result()
lines = [line.strip() for line in texts.strip().split("\n") if line.strip()]
if not lines:
return "No valid texts found.
"
start = time.time()
all_results = classifier(lines)
elapsed = (time.time() - start) * 1000
counts = {"Positive": 0, "Negative": 0, "Neutral": 0}
html_parts = []
for text, results in zip(lines, all_results):
top = results[0]
label = top["label"]
score = top["score"]
counts[label] = counts.get(label, 0) + 1
cfg = SENTIMENT_CONFIG.get(label, SENTIMENT_CONFIG["Neutral"])
bar_width = int(score * 100)
html_parts.append(f"""
đš FinBERT-Multi
Financial sentiment analysis powered by FinBERT, fine-tuned on 143K+ samples from 5 expert datasets
đ§ ENTUM-AI / FinBERT-Multi
đ 143K+ training samples
""")
with gr.Tabs():
# --- Tab 1: Single Analysis ---
with gr.Tab("đ Single Analysis"):
with gr.Row():
with gr.Column(scale=3):
single_input = gr.Textbox(
label="Financial Text",
placeholder="e.g. Stock price soars on record-breaking earnings report",
lines=2,
max_lines=4,
)
single_btn = gr.Button("⥠Analyze Sentiment", variant="primary", size="lg")
with gr.Column(scale=4):
single_output = gr.HTML(value=create_empty_result())
gr.Examples(
examples=[
["Stock price soars on record-breaking earnings report"],
["Revenue decline signals weakening market position"],
["Company announces quarterly earnings results"],
["Shares surge 15% after strong Q3 revenue growth"],
["Major layoffs expected as company restructures operations"],
["The board of directors met to discuss routine operations"],
["Bankruptcy filing raises concerns about long-term viability"],
["Profit margins improved significantly driven by cost optimization"],
],
inputs=single_input,
label="đ Try these examples",
)
single_btn.click(fn=predict_single, inputs=single_input, outputs=single_output)
single_input.submit(fn=predict_single, inputs=single_input, outputs=single_output)
# --- Tab 2: Batch Analysis ---
with gr.Tab("đ Batch Analysis"):
gr.Markdown("Paste multiple financial texts â **one per line** â for batch sentiment classification.")
with gr.Row():
with gr.Column(scale=2):
batch_input = gr.Textbox(
label="Financial Texts (one per line)",
placeholder="Headline 1\nHeadline 2\nHeadline 3",
lines=8,
max_lines=20,
)
batch_btn = gr.Button("⥠Analyze All", variant="primary", size="lg")
with gr.Column(scale=3):
batch_output = gr.HTML(
value="Results will appear here.
"
)
batch_btn.click(fn=predict_batch, inputs=batch_input, outputs=batch_output)
# --- Tab 3: About ---
with gr.Tab("âšī¸ About"):
gr.HTML("""
About FinBERT-Multi
A financial sentiment model built on
ProsusAI/FinBERT.
Fine-tuned on 143K+ samples from 5 combined financial datasets for maximum coverage and robustness.
| Base Model |
ProsusAI/FinBERT (BERT-based) |
| Task |
3-Class Sentiment (Positive / Negative / Neutral) |
| Training Data |
143K+ samples from 5 datasets |
| Language |
English |
| License |
Apache 2.0 |
đ Training Datasets
| Dataset |
Samples |
| FinanceInc/auditor_sentiment | ~4.8K |
| nickmuchi/financial-classification | ~5K |
| warwickai/financial_phrasebank_mirror | ~4.8K |
| NOSIBLE/financial-sentiment | ~100K |
| TimKoornstra/financial-tweets-sentiment | ~38K |
đ Python API
from transformers import pipeline
classifier = pipeline("text-classification",
model="ENTUM-AI/FinBERT-Multi")
result = classifier("Stock price soars on earnings")
# [{'label': 'Positive', 'score': 0.99}]
""")
# Launch
demo.launch()