13-B | Devpost

chat para preguntar sobre cualquier producto o categoria
categoria mostrando sus productos
Palabras positivas y negativas mas repetidas
palabras mas repetidas en todas las reseñas en un producto
resumen positivo y negativo hecho en base a reviews de clientes. Dice especificamente porque les gusto o no les agrado el producto

Inspiration

The e-commerce industry is growing rapidly. Along with it, the amount of information that the customer and user are providing as feedback is increasing. Ecommerce owners and entrepreneurs are struggling to analyze their customers' comments and reviews due to the same reason: too much information and data. In that large amount of information, a large percentage is non-useful text and most of the time valuable data is lost or difficult to find in the pile of text. Detailed analyzes cannot be reached and instead they are more general.

What it does

Automation with its own LLM model in the analysis of customer feedback about a product. Our model eliminates junk text and uses specific key words and phrases for better analysis. Frequency identifier of positive and negative words. Identify the most used words. Summary by positive and negative section using these frequencies.

How we built it

Using the OpenAI API documentation we created an LLM model in which we entered 19,662 product reviews to create its knowledge. There is a Chatbot where you can ask the model any questions about the products, categories, reviews, etc. Based on this, we divided the products that were within these reviews into categories and then entered the products into their respective category. When entering a product, the model analyzes all the words in the reviews that that product has and divides the information from all the reviews as follows:

percentage of customer happiness in the product
percentage of customer displeasure in the product
most used positive words
most used negative words
positive summary: the model analyzes all the reviews and creates a negative summary using the most important and used negative words and phrases, creating a more detailed analysis of why the customer did not like it
negative summary: the model analyzes all the reviews and creates a positive summary using the most important and used positive words and phrases, creating a more detailed analysis of why the customer liked it

Challenges we ran into

The deployment of python functions in the cloud as well as learning how to create an LLM model with our own data. Likewise, being able to "pass" data to the user interface.

Accomplishments that we're proud of

We got the model to analyze a large number of reviews by product or categories and thus be able to obtain the key words and phrases that the client entered as a review. We managed to count the words that marked

What we learned

We learned how to create our own LLM model where we gave our data to use. In addition, it could analyze large amounts of reviews, determine and count the most important words and phrases in all the reviews within a product.