Project 2: Movie Sentiment Analysis

This project helps you move beyond toy classification by handling noisy text and interpretation of model behavior.

Problem

Predict whether a movie review is positive or negative.

Baseline Pipeline

from sklearn.pipeline import Pipeline
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression

pipe = Pipeline([
    ("tfidf", TfidfVectorizer(stop_words="english", max_features=50000)),
    ("model", LogisticRegression(max_iter=3000))
])

pipe.fit(X_train, y_train)

Evaluation Checklist

• accuracy for quick baseline
• precision/recall/F1 for class-wise quality
• confusion matrix
• inspect top false positives/false negatives

Real Example Analysis

Common false negative:

• review contains sarcasm ("great... if you like wasting time")

Improvement ideas:

• add bi-grams/tri-grams
• compare with linear SVM
• perform targeted text normalization

Deliverables

1. Baseline model notebook
2. Comparison table of at least 2 models
3. Error analysis section with 10 misclassified reviews