Restaurant Recommendation System using Aspect based Sentiment Analysis

With the advent of the digital universe, e-commerce, and social networks, the Web now offers a wide range of information, products, and services. A recommender system is a type of information filtering technology that can be used to forecast ratings for items (such as products, services, and movies)

Project Title

Project Area of Specialization

Artificial Intelligence

Project Summary

Recommender systems are being used in the business of restaurants, making use of parameters, such as ‘star ratings’. This project introduces a novel approach of recommendation for restaurants using user reviews available on review websites. Restaurant reviews will be scrapped from review websites like TripAdvisor. Extracted reviews will be stored in a relational database. These reviews will be sentimentally analyzed and scored. For each query, the system will extract aspects present and understand the semantics of the question. These aspects and the context-aware details will be searched for, in the stored reviews. Based on the scores, derived from considering various factors relevant to the domain, we will provide a list of recommendations to the user. This system will serve as an authentic review and recommendation hub for food and travel enthusiasts.

Project Objectives

The product here is a recommendation system, using advanced natural language processing techniques such as NER and sentiment analysis to recommend a restaurant to the customer based on their likings. Users can ask a question or write a statement relevant to the food and restaurant domain. The system will pre-process the query and deal with it accordingly, providing personalized recommendations.

Project Implementation Method

To develop this system, we will be using Python3. To scrap reviews and data, Selenium and Beautiful Soup are used. For NER (named entity recognition), the SpaCy v3 library is used. For the front-end, we will be using Bootstrap, HTML, and CSS. For the backend, we will be using the Django framework.

Benefits of the Project

Benefits of the Project:

1. No constraint on choice. Users can ask about anything domain-specific to restaurants, without any limits on filters (pre-defined tags).

2. In comparison with previous research and implementations, the proposed system covers almost every aspect of the restaurant domain; whereas previous systems are limited to food, ambience and service preferences.

3. Most of the time, we want to try out something new. The proposed system doesn’t need the user to have a preference. It will provide word-of-mouth references or recommendations from a huge number of people.

4. The proposed system will work with a fine-grain aspect and semantic extraction model to understand what is desired and search about by the user.

5. Overcomes the limitation of star/bubble rating system.

Technical Details of Final Deliverable

The list and details of final deliverables are:

1. Scrappers (to scrap data from websites such as Trip Advisor, Yelp, Zomato etc.)

2. Annotation Tool (To create a aspect-based sentiment analysis)

3. Food Ontology (to recognize any food or drink into a specific category)

4. Deep Learning Model trained to Recommend (this will be a transformer-based deep learning model that will recommend based on publically available user review data on restaurants)