Sentiment Analysis

Sentiment analysis (also called opinion mining) is the task of determining the emotional tone or attitude expressed in a piece of text. It is one of the most popular and commercially valuable applications of NLP.

---

Types of Sentiment Analysis

Type	Description	Output Example
Polarity detection	Positive, negative, or neutral	"Great movie!" → Positive
Fine-grained	Star rating or scale (1–5)	"Decent but not amazing" → 3 stars
Aspect-based	Sentiment per aspect of an entity	"Great food but slow service" → food: +, service: -
Emotion detection	Specific emotions (joy, anger, sadness, etc.)	"I'm furious about this!" → Anger

---

Challenges in Sentiment Analysis

Challenge	Example
Sarcasm	"Oh wonderful, my flight is delayed again" → Negative (not positive)
Negation	"This is not a bad product" → Positive (double negative)
Context dependence	"This phone's battery lasts forever" (positive) vs "This meeting lasts forever" (negative)
Mixed sentiment	"The camera is amazing but the battery life is terrible"
Domain specificity	"Unpredictable plot" → Positive (for a thriller), Negative (for a manual)
Implicit sentiment	"I waited two hours for my order" → Negative (no explicit sentiment words)

---

Rule-Based Approaches

VADER (Valence Aware Dictionary and sEntiment Reasoner)

VADER is a rule-based sentiment analyser specifically tuned for social media text. It handles emojis, slang, capitalisation, and punctuation.

import nltk
nltk.download('vader_lexicon')
from nltk.sentiment import SentimentIntensityAnalyzer

sia = SentimentIntensityAnalyzer()

texts = [
    "I love this product! It's absolutely amazing!",
    "This is the worst experience I've ever had.",
    "The movie was okay, nothing special.",
    "GREAT job!!! :D",
]

for text in texts:
    scores = sia.polarity_scores(text)
    print(f"{text}")
    print(f"  Scores: {scores}")
    print()

VADER output format:

Score	Meaning
neg	Proportion of negative sentiment
neu	Proportion of neutral sentiment
pos	Proportion of positive sentiment
compound	Overall sentiment score (-1 to +1)

Tip: The compound score is the most useful. Typically: compound >= 0.05 → Positive, compound <= -0.05 → Negative, otherwise → Neutral.

TextBlob

from textblob import TextBlob

text = "The food was delicious but the service was terrible."
blob = TextBlob(text)
print(f"Polarity: {blob.sentiment.polarity:.2f}")    # -1 to 1
print(f"Subjectivity: {blob.sentiment.subjectivity:.2f}")  # 0 to 1

---

Machine Learning Approaches

TF-IDF + Classifier

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report