If you have new seasonal and longtail products in your catalogue, chances are they don’t get picked up by the recommendation engine. Traditional recommendation engines are based on historical co-occurrence, that is, they should have been bought together or viewed together in the past. Due to this, certain products remain buried, and shoppers cannot find them. We solve this constraint by using NLP or natural language processing. It works by looking at a product description and comparing these words with other items that have a similar description.
Let me demonstrate how Saxo, a bookseller in Denmark is using NLP for better product discovery. I have chosen a cookbook on the website and you can see the product description here similar to what you’d find on the back cover. What NLP is going to be doing is picking out words such as slow cooker, crock pot, and looking for similar descriptive text in other products. So you first see a section that is using a co-occurrence algorithm to show other books that are usually viewed together. Underneath here, we see the section called books on the same subject. And this is based on the NLP algorithm that automatically inferred the relationships between these products based on the wealth of text descriptions and other data sources.
Let me go to this other slow cooking book and I see text that’s very similar in meaning to the original book I was viewing. The second thing Saxo is using NLP for is to find books by the same author. Here I am looking at a book by Stephen King. And as I scroll down, there is a section with books that others also looked at. Then there is the RichRelevance(now Algonomy) NLP widget for other books from the same author. And that makes product discovery easy for me while helping Saxo to surface products that have no past interactions or behavioural data. This innovation is a game changer for businesses with a fast changing catalogue and helps improve conversions by providing relevant recommendations to shoppers as they browse the website.
