How To Extract Unique Features From User-Generated Content

ARTIFICIAL INTELLIGENCE IN ECOMMERCE - PART 2

What kind of a jacket would you need for a hiking the coast of Oregon where conditions change in a matter of minutes?

Ask any hiking enthusiast sales associate in REI or a Columbia store in Oregon and they will happily help you buy the right jacket. Easy right. Not so much if you are online shopping for this special jacket from the comfort of your couch.  What would you look for? the wind protection ability OR ability to protect from light horizontal rain /ocean spray OR a jacket which does not get too warm while walking OR a jacket which is light enough to carry in case it turns out to be sunny OR maybe all of the above.

When you are buying specialized merchandise, user-generated content is the best place to look. You can spend minutes or hours or days researching the weight, material, uses, features, etc for each jacket until you find the right one.

Now imagine a system that can read your mind as you go about your business on the website and somehow present you the correct information from product descriptions, product reviews, and product Q& A, all especially contextualized for you.  A shopper buying a lightweight rain jacket is less concerned if a jacket was shipped on time or if the package was damaged. They want to read reviews about what matters most to them. The weight and material.

From a merchant's point of view, the issue is even more critical. More time a shopper spends researching and making a buying decision lesser the chances for a purchase.  So it's extremely critical for them to then quickly understand the shopper's intent, identify minimal information for effective decision making and present it in the most efficient manner.

User-generated content can present a wealth of decision-making information such as subjective feature descriptions and unique product usage. However, mining this information along with the sentiment remains challenging. This is one area where artificial intelligence tools can excel and improve the shopper experience many folds. Trained natural language processing models can be put in place to scan product reviews just like a human shopper would and then match the shopper's intent to statements in the product reviews which depict strong sentiment for that intent.

This approach differentiates each product from another in the context of the shopper while highlighting the minimal most relevant information to the shopper. The insights extracted from user-generated content can also be used to fill gaps in product descriptions, product search keywords and product features lists.

This post again is one of many ways where intelligent machine learning algorithms combined with data can make the experience much more powerful and engaging. No more ugly tag clouds generate or a long list of reviews thrown at the shopper to figure out.