Oreoreo

Inspiration

Online reviews are a critical source of information for consumers, but their value is diminished by the presence of irrelevant, misleading or low-quality content. This project addresses the core challenge of automatically assessing the quality and relevance of Google location reviews using a machine learning-based system. Our goal is to develop a robust pipeline that can effectively detect and classify reviews that violate a set of predefined policies, specifically identifying advertisements, irrelevant content and rants. By doing so, our solution aims to enhance the reliability of review platforms, improve user trust and ensure fair representation for businesses.

What it does

This project processes, analyses and classifies online reviews. It combines natural language processing, feature engineering and machine learning to indentify three types of reviews:

Normal - Genuine reviews Advertising - Promotional or marketing content Irrelevant content - Off-topic or vague, unhelpful reviews

How we built it

The pipeline includes sentiment analysis, TF-IDF, sentence embeddings, rating-category alignment and Random Forest classification with synthetic data augmentation.

Challenges we ran into

Being just year 2 Business Analytics students, there was alot of content that we had to learn on our own about the use of machine learning as well as the necessary conditions needed before running the code to help us check for reviews.

Accomplishments that we're proud of

Being able to come up with a working model under the time crunch that actually could output a successful solution.

What we learned

We learnt and touched the basics of machine learning, which we are all very keen to dive further into, which also will help us in our university journey

What's next for Oreoreo

Continuing with more hackathons to improve our skills and gain more exposure into how the things we are learning now will affect the real working world.