In today's information age, product reviews play an important role in users' purchasing decisions.
For this reason, this project focuses on automating the extraction of keywords from reviews of e-commerce portals. To achieve this, the Yelp portal was chosen as a source of user reviews.
Automation will be achieved through the use of different methods from one of the most popular fields of artificial intelligence, namely NLP - Natural Language Processing -.
The aim of this work is to try three different classical techniques in order to extract keywords. After that, we are going to use a BERT approach and compare the results that were obtained.
-
TF-IDF
-
LDA
-
Text Rank
-
BERT + TF-IDF
-
BERT + LDA
-
BERT + Text Rank
The data is located in the folder 'data'. We are going to prove our approach with a bunch of reviews which belongs to important business. Such as Mcdonalds, Starbucks and CVS Pharmacy.