Skip to content
Debuggerpaco
ProjectsNLP / Machine Learning

PROJECT 03 / 06

Hotel Reviews Sentiment Analysis

A notebook-based comparison of classical machine learning, LSTM, and DistilBERT sentiment approaches on hotel reviews.

View source on GitHub
Focus
NLP / Machine Learning
Tools & methods
Python · TF-IDF · LSTM · DistilBERT · Jupyter
Illustrated Hotel Reviews Sentiment Analysis cover with a hotel and sentiment dashboard
Illustrated Hotel Reviews Sentiment Analysis cover with a hotel and sentiment dashboard

01

The question

How do different sentiment approaches behave on the positive and negative text fields of the same hotel-review dataset? This project uses 1,500 La Veranda Hotel reviews from a Booking.com dataset published on Kaggle.

02

The pipeline

  1. Clean the review data and prepare separate positive-review and negative-review tracks.
  2. Derive sentiment labels from review scores: 1-4 negative, 5-7 neutral, and 8-10 positive.
  3. Compare TF-IDF features with six classical machine-learning classifiers, then train LSTM models for both tracks.
  4. Run DistilBERT zero-shot sentiment analysis and compare approaches within the notebook workflow.

03

What is available

The repository contains four ordered notebooks covering preprocessing, classical models, deep learning, and DistilBERT comparison, together with setup requirements and a link to the source dataset. The portfolio cover is an illustration, not a deployed application dashboard.

04

Interpretation and limits

Score-derived labels are a proxy for sentiment, not independently annotated ground truth. A single-hotel dataset also limits generalization to other hotels, languages, and review styles. No accuracy claim or hosted-demo claim is made here without a documented evaluation.

Explore the work

Source code, documentation, and reproducibility notes on GitHub