Create an IMDB Sentiment Classifier with Scikit‑LLM & Groq in 500 Lines

Prerequisites and Environment Setup

Install the Scikit‑LLM library and configure it to use Groq’s OpenAI‑compatible endpoint.

pip install scikit-llm

from skllm.config import SKLLMConfig
SKLLMConfig.set_gpt_url("https://api.groq.com/openai/v1")
SKLLMConfig.set_openai_key("YOUR-API-KEY")

Data Acquisition

Load the public IMDB sentiment CSV from GitHub, then sample 500 rows for a lightweight demo.

import pandas as pd
from sklearn.model_selection import train_test_split
url = "https://raw.githubusercontent.com/Ankit152/IMDB-sentiment-analysis/master/IMDB-Dataset.csv"
df = pd.read_csv(url)
df_sampled = df.sample(n=500, random_state=42)
X = df_sampled["review"]
y = df_sampled["sentiment"]  # 'positive' or 'negative'
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42)

Text‑Cleaning Transformer

Define a function that removes HTML tags and collapses whitespace, then wrap it with FunctionTransformer so it can be used inside a scikit‑learn pipeline.

from sklearn.preprocessing import FunctionTransformer

def clean_text_data(texts):
    series = pd.Series(texts).astype(str)
    cleaned = series.str.replace(r'<[^>]+>', ' ', regex=True)  # strip HTML tags
    cleaned = cleaned.str.strip().str.replace(r'\s+', ' ', regex=True)  # collapse spaces
    return cleaned.tolist()

text_cleaner = FunctionTransformer(clean_text_data)

Sentiment‑Analysis Pipeline

Chain the cleaner with a zero‑shot LLM classifier that calls Groq’s Llama 3.1 8B model. The fit call does not train the model; it only registers the two class labels present in y_train.

from sklearn.pipeline import Pipeline
from skllm.models.gpt.classification.zero_shot import ZeroShotGPTClassifier

sentiment_pipeline = Pipeline([
    ("cleaner", text_cleaner),
    ("llm_classifier", ZeroShotGPTClassifier(model="custom_url::llama-3.1-8b-instant"))
])

print("Fitting the pipeline...")
sentiment_pipeline.fit(X_train, y_train)

Evaluation

Run predictions on the held‑out test set and compute a scikit‑learn classification report. The sample loop prints three truncated reviews with their true and predicted labels.

from sklearn.metrics import classification_report

predictions = sentiment_pipeline.predict(X_test)
print(classification_report(y_test, predictions))

for review, actual, pred in zip(X_test[:3], y_test[:3], predictions[:3]):
    short = review[:100] + "..."
    print(f"Review: {short}")
    print(f"Actual: {actual} | Predicted: {pred}")

--- Classification Report ---
              precision    recall  f1-score   support
    negative       0.95      0.97      0.96        60
    positive       0.95      0.93      0.94        40
    accuracy                               0.95       100
   macro avg       0.95      0.95      0.95       100
weighted avg       0.95      0.95      0.95       100
--- Sample Predictions ---
Review: I saw mommy...well, she wasn't exactly kissing Santa Clause; he has his hand on her thigh and wicked...
Actual: negative | Predicted: negative
Review: This entry is certainly interesting for series fans (like myself), but yet it is mostly incomprehens...
Actual: negative | Predicted: negative
Review: Ingrid Bergman (Cleo Dulaine) has never been so beautiful. Gary Cooper as "Cleent" so perfectly cast...
Actual: positive | Predicted: positive

The pipeline achieves approximately 0.95 accuracy on the 100‑sample test split, demonstrating that a classic scikit‑learn workflow can be extended to zero‑shot LLM‑driven sentiment classification with minimal code.