Scentalytics
Sentiment Analysis
Consumer Feedback / Likes / Dislikes
About
Sentiment analysis is a very popular method of discovering what your customers think about your products. There are different sentiment analysis techniques available today, from simple customer surveys to advanced machine learning models which understand customer reviews written in natural language.
Such models can say whether a review is positive, negative, or neutral, or does it even belong to one of the more detailed categories of sentiment. There are also models which can predict a numeric score instead of a category.
The data about perfumes which Scentalytics analyzed contains sentiment information in form of percentages of consumer votes for five different categories, from very positive to very negative. This information allows us to not only see which products are liked the most, but even brands, perfumers or perfume types. Check out how we used that data to analyze sentiments of fragrances from Tom Ford and Kilian Paris.
Our AI Perfume Creator contains two sentiment analysis models - one classifier for different categories and one regression model which can predict the potential numerical rating for the given perfume composition. We can also discover which other factors, such as notes, longevity or sillage, have the highest impact on the sentiment scores.
Challenge
Answer questions like:
-
Which products, brands, perfume notes and accords are most loved by consumers?
-
How to predict a success of a perfume product?
Solution
We created top lists of perfumes, brands, perfumers and combinations of notes which consumers like the most. We also developed cusotm machine learning models for sentiment analysis which can be used to assess the potential sentiment of future products.
Results
The result of our sentiment analysis are top lists and bar charts of the most loved perfume products, brands, perfumers, perfume types and notes. Our AI models for sentiment analysis can predict the average sentiment scores and ratings for a future product, based on perfume notes and accords.
6
Sentiment categories
1,900+
Notes and accords analyzed
71,000+
Perfumes analyzed
