Kewlkaka/customer-complaint-NLP-text-classification ? reverse-engineered prompt

Build me a notebook that classifies consumer complaint text into the right product category, like credit card, mortgage, debt collection, and student loan.

I want the full workflow in one place, from cleaning the complaint text to training and comparing a few classic machine learning models. Please include text preprocessing like lowercasing, removing punctuation and numbers, stop word removal, and lemmatization, then turn the text into features with TF IDF using unigrams and bigrams. It should also handle class imbalance by balancing the data and doing simple augmentation for smaller groups if needed.

Train and evaluate models like SVM, Naive Bayes, Logistic Regression, and a couple of others if useful, then tune the best one with grid search. Show the accuracy and a clear classification report so I can see how each category performs. If you need anything extra from the latest docs, look it up online and keep the notebook easy to run from start to finish.