Restaurant reviews







Restaurants consider reviews important because of how they can help them improve their customer satisfaction levels. By knowing what customers like about their business, they can improve on areas of strength. Knowing what customers dislike will help them identify where to invest more time and money to correct those problems.



The problem

By taking on classification problems, humans can understand the world better. The application of this technique in the real-world is a rather more challenging task because of the fine lines between classes. However, AI makes classification despite these fine lines, easier and faster.

In this challenge, we will develop a model to categorize restaurant reviews into positive or negative reviews. Doing this is beneficial to not only the restaurants but also the customers so they can make the right decisions on restaurant choice, menu selection, etc.

The reputation and success of a restaurant are invariably tied to the experience they can offer their customers. How much, therefore, does the nature of customer reviews impact the business? 


what is classification?

Although you will learn more about classification in the second course of Level2 Supervised Learning, in the AI Master Program, here’s a brief idea: classification is a technique under supervised learning in which a model takes in an input (here, satellite images), learns from the input and then proceeds to classify any new input it is fed with.


learning outcomes

  • From the knowledge you gain from this project, you will grasp how (to):

    • Gather data and divide it into training and testing datasets.
    • Create an ML model to solve a classification problem of splitting restaurant reviews into positive/negative comments.  


The data

The dataset required for this project is a set of restaurant reviews that are categorized as positive/negative.

The dataset will train the ML model to classify reviews on its own. To ascertain the performance of your model, be sure to split the data into testing data and training data.  


PROJECTS STEPS: 1. Meet your ai identity

Meet your AI Identity. You will teach your AI how to complete tasks as a human would. In due time, your AI Identity will grow more intelligent as you train it some more to take on different tasks with multiple projects. Eventually -in the future- you will be able to transfer your AI identity's intellect into a physical robot you can interact with in the real world.  



Begin every project by naming the project so you can keep track of what project it is exactly. Every project involves the steps of dataset preprocessing, model building, evaluation, and deployment. By creating projects, working with others becomes more convenient and easier.

Click CREATE once you have added a project name and description. 



Here, from the options of a Computer Vision model (MAKE ME SEE), Natural Language Processing model (MAKE ME READ), Voice Recognition (MAKE ME HEAR), and a tabular model (MAKE ME COUNT), you will select MAKE ME READ.

This model will enable you to train your AI transformer to recognize and classify the restaurant reviews.

Keep track of the newer models that will be added to the platform. Each one of these models will enable your AI transformer to take on and solve new problems.

More AI capabilities (models) will be added to the platform, so stay tuned! These models will help you to accelerate the process of creating a trained AI model that solves your problem, and thus export into your app to start using it!



Use either the “Restaurant Reviews” dataset from the dataset library or manually sourced reviews from recognized sources such as Kaggle to collect the reviews you need for this project.

Take note that data gathering is a very crucial stage of an ML project with data preprocessing taking at least 70% of the time. Therefore, you will need to carry out this stage well.

Once you have the questions, classify the data as follows: “Positive” or “Negative”. 



Use the reviews you have gathered to train your AI transformer. “Training” your model refers to teaching it to recognize categorize reviews. The aim of this is so the AI transformer can categorize new reviews into the corresponding category. To do this, click TRAIN ME. Be patient for a moment while this process goes on.




With your AI model trained now, it is time to test the model by feeding it new reviews. In this testing phase, take the precaution of not using any review that was used during the training phase.

The PREVIEW section will serve to determine how well the model has been trained. All you have to do is upload/import the new reviews to this section. The confidence score will help you know well the model has been trained and if you are fine with it, export the model.

If you do not like the confidence score, return to step 4, and determine whether enough reviews were used to train the model. 




Exporting your model is the next step after you are satisfied with the confidence score. You can export using our API into your website, application, robot, or any other machine.

Return to step 4 if you got a low confidence score. Doing this helps you review the training data class to be sure it is of high quality (clean) and quantity.

Your model is only as good as the data you feed it.