NLP_ASSIG.py

"""
NLP Short Assignment – Sentiment Analysis
-----------------------------------------
1. Dataset Prep
2. Prompt Engineering
3. Evaluation
4. Troubleshooting
"""

# =============================
# 1) NLP & Dataset Prep
# =============================
import re

# Mini IMDb-style dataset (public-domain review style, small for demo)
DATASET = [
    {"text": "Absolutely loved it. The performances were outstanding!", "label": 1},
    {"text": "Terrible movie. Boring plot and a waste of time.", "label": 0},
    {"text": "What a delightful surprise! Smart writing and heartfelt moments.", "label": 1},
    {"text": "I wanted to like it, but it was not good. Confusing and slow.", "label": 0},
    {"text": "Great soundtrack and visuals. I had a great time watching it!", "label": 1},
    {"text": "This is bad. The jokes never land and the pacing is awful.", "label": 0},
]

# Cleaning + tokenization
def clean_text(text: str) -> str:
    text = text.lower()
    text = re.sub(r"[^a-z0-9\s]", "", text)
    return text.strip()

def tokenize(text: str):
    return clean_text(text).split()

print("Sample cleaned review:", clean_text(DATASET[0]["text"]))

# =============================
# 2) Prompt Engineering & Model
# =============================
try:
    from transformers import pipeline
    clf = pipeline("sentiment-analysis",
                   model="distilbert-base-uncased-finetuned-sst-2-english")
    use_llm = True
except Exception:
    # Fallback heuristic classifier if transformers not available
    POS = {"love", "great", "wonderful", "delightful", "smart", "outstanding"}
    NEG = {"terrible", "boring", "bad", "awful", "waste", "slow"}
    
    def clf(text):
        tokens = tokenize(text)
        score = sum(1 for t in tokens if t in POS) - sum(1 for t in tokens if t in NEG)
        return [{"label": "POSITIVE" if score >= 0 else "NEGATIVE",
                 "score": abs(score) / 3}]
    
    use_llm = False

# Example text for prompts
sample_text = DATASET[0]["text"]

# Three prompt variations
prompts = [
    f"Classify the sentiment of this review: {sample_text}",
    f"Is this movie review positive or negative? Review: {sample_text}",
    f"Analyze and return JSON with 'label' and 'confidence'. Text: {sample_text}",
]

print("\n--- Prompt Engineering Outputs ---")
for i, p in enumerate(prompts, 1):
    out = clf(sample_text)[0]
    if i == 3:
        print(f"Prompt {i} -> {{'label': '{out['label']}', 'confidence': {out['score']:.2f}}}")
    else:
        print(f"Prompt {i} -> {out['label']} (conf {out['score']:.2f})")

# =============================
# 3) Evaluation
# =============================
y_true = [d["label"] for d in DATASET]
y_pred = [1 if clf(d["text"])[0]["label"] == "POSITIVE" else 0 for d in DATASET]

# Metrics (pure python version)
def compute_metrics(y_true, y_pred):
    tp = sum(t == p == 1 for t, p in zip(y_true, y_pred))
    tn = sum(t == p == 0 for t, p in zip(y_true, y_pred))
    fp = sum(t == 0 and p == 1 for t, p in zip(y_true, y_pred))
    fn = sum(t == 1 and p == 0 for t, p in zip(y_true, y_pred))
    acc = (tp + tn) / len(y_true)
    prec = tp / (tp + fp) if tp + fp else 0
    rec = tp / (tp + fn) if tp + fn else 0
    f1 = 2 * prec * rec / (prec + rec) if prec + rec else 0
    return acc, prec, rec, f1

acc, prec, rec, f1 = compute_metrics(y_true, y_pred)
print("\n--- Evaluation Metrics ---")
print(f"Accuracy: {acc:.2f}, Precision: {prec:.2f}, Recall: {rec:.2f}, F1: {f1:.2f}")

# =============================
# 4) Troubleshooting
# =============================
print("\n--- Troubleshooting Note ---")
print("Issue: Sarcasm and negation (e.g., 'yeah right, great movie') confuse models.")
print("Fix: Add sarcastic examples to dataset, or prompt model to consider sarcasm explicitly.")

# =============================
# Sanity Tests
# =============================
def run_tests():
    assert clean_text(" Hello!! ") == "hello"
    assert tokenize("Not great!!") == ["not", "great"]
    print("All tests passed ✅")

run_tests()