Opportunities: What Terms Are Included In Word Clouds, And How Are They Sized?

Word clouds within Assortment Opportunities are generated using classical machine learning to identify the most relevant words from the name, brand, and description fields within a group of products.

For example, if we’re looking at Men’s T-shirts, the word “T-shirt” will frequently appear in the product names across the filter set. However, the term “T-shirt” is unlikely to be helpful when analyzing product attributes.

Once we have a list of relevant terms, the size of each term within the word cloud is dictated by its relevance across the whole filter set.

We do this by comparing the term's frequency within the Opportunity with the term's average use across the whole filter set. We reduce the values of common terms and boost the values of less common terms that are distinctive for the specific Opportunity presented.