Development of a product review website powered by an NLP model(s) The model(s) aggregates customer feedback from different sources Find and reuse existing LLMs, don’t use classical techniques Main tasks: Classify customer reviews into positive, negative, or neutral categories to help the company improve its products and services. Cluster product categories into just 4-6 of them Use Generative AI to summarize reviews into an article which recommends the top products for each category
Create a model for classification of customers' reviews (the textual content of the reviews)into positive, neutral, or negative.
The dataset categories are too sparse. Create a model to group all reviews into 4-6 “meta” categories.
Create a model which, for each product category, generates a short article, like a blogpost reviewer would write, to help customers choose the best one for them.
This project was quite challenging. I encountered several problems and spent a significant amount of time working with unfamiliar techniques. As the deadline approached, I simplified my approach to save time. While I achieved results that were acceptable and aligned with my goals for the summarization model, there is considerable room for improvement. Specifically, I aim to enhance the training part of the code and refine the clustering process to achieve better results.
https://solvedbrunus.blogspot.com/2024/12/how-transformers-revolutionize.html