Add TF-IDF + LinearSVC TweetEval sentiment example tuned with Optuna (with smoke test & README)

[aish-warya-iyer]

This example adds a short, self-contained demonstration of using Optuna to tune a TF-IDF + LinearSVC pipeline for sentiment classification on the TweetEval sentiment dataset (three labels: negative, neutral, positive).

What it does

Loads the TweetEval sentiment dataset using the datasets library.

Tunes both the TF-IDF vectorizer (feature count, n-gram range, etc.) and LinearSVC parameters (C, loss, class_weight) using Optuna.

Uses macro-F1 on the validation split as the objective (1 – macro-F1 minimized).

Retrains the best configuration on train + validation and prints a test report.

Files added

examples/sklearn/svm_tfidf_tweeteval_sentiment.py – main script

examples/sklearn/svm_tfidf_tweeteval_sentiment.md – short usage notes

tests/test_svm_tfidf_tweeteval.py – quick smoke test for CI

How to run
python examples/sklearn/svm_tfidf_tweeteval_sentiment.py --n-trials 20 --max-train 20000
pytest -q # optional quick test

Notes

Keeps runtime light by allowing the --max-train argument to limit samples.

Demonstrates how Optuna can help search SVM + text-feature spaces efficiently.

No external dependencies beyond datasets, scikit-learn, and optuna.

[github-actions]

This pull request has not seen any recent activity.

[github-actions]

This pull request was closed automatically because it had not seen any recent activity. If you want to discuss it, you can reopen it freely.

[c-bata]

@aish-warya-iyer Please feel free to reopen this after fixing all CI checks 🙏