Background: Product reviews in the form of text snippets abound online -- from Google restaurant reviews, to Amazon product reviews, Rotten Tomato movie reviews, and anonymous comment threads on Reddit. Understanding the sentiment of these text reviews can provide insights into customer satisfaction, and therefore consumer demand, for the products / services that are not formally reviewed.

Objective: Use natural language processing techniques (neural networks with sentence/text embedding layers) to identify sentiment vectors, which can then be used to predict review rating, and/or identify similar products.

Goal: Predict customer reviews (stars) based on written review text.

Data: Google Local reviews