ONES-RS Benchmarks

-- Snowflake Cortex Sentiment Analysis
SELECT
    FULL_SENTENCE,
    SNOWFLAKE.CORTEX.SENTIMENT(FULL_SENTENCE) AS sentiment_score,
    CASE
        WHEN sentiment_score >= 0.60 THEN 'STRONG_POSITIVE'
        WHEN sentiment_score >= 0.25 THEN 'POSITIVE'
        WHEN sentiment_score > -0.25 AND sentiment_score < 0.25 THEN 'NEUTRAL'
        WHEN sentiment_score <= -0.25 THEN 'NEGATIVE'
        ELSE 'STRONG_NEGATIVE'
    END AS sentiment_label
FROM FINANCIAL_PHRASEBANK_SENTENCES;

from ones_rs.ones_rs import OnesEngine
from sklearn.metrics import accuracy_score, f1_score

# Initialize ONES-RS with financial domain lexicon
engine = OnesEngine()
engine.load_lexicon('financial_lexicon.json')

# Analyze all sentences
results = engine.analyze_batch_auto(sentences)
predictions = [r[2] for r in results]  # sentiment label

# Calculate metrics
accuracy = accuracy_score(ground_truth, predictions)
f1_macro = f1_score(ground_truth, predictions, average='macro')

import pandas as pd
from sklearn.metrics import accuracy_score, f1_score, classification_report
from ones_rs.ones_rs import OnesEngine

# Load Snowflake export
df = pd.read_csv('snowflake_export.csv')

# Extract ground truth from "sentence@label" format
df['ground_truth'] = df['FULL_SENTENCE'].apply(
    lambda x: x.rsplit('@', 1)[1].strip() if '@' in str(x) else None
)
df = df[df['ground_truth'].notna()]

# Map Snowflake labels to 3-class
def map_label(label):
    label = str(label).lower()
    if 'positive' in label: return 'positive'
    elif 'negative' in label: return 'negative'
    return 'neutral'

df['snowflake_pred'] = df['SENTIMENT_LABEL'].apply(map_label)

# Run ONES-RS
engine = OnesEngine()
engine.load_lexicon('financial_lexicon.json')
sentences = [s.rsplit('@', 1)[0] for s in df['FULL_SENTENCE']]
results = engine.analyze_batch_auto(sentences)
df['ones_pred'] = [r[2] for r in results]

# Compare
print("Snowflake Accuracy:", accuracy_score(df['ground_truth'], df['snowflake_pred']))
print("ONES-RS Accuracy:", accuracy_score(df['ground_truth'], df['ones_pred']))
print("\nONES-RS Classification Report:")
print(classification_report(df['ground_truth'], df['ones_pred']))