The idea of Natural Language is the act of having a conversation with your computer in the same way you talk to your friends of family.  It’s a conversation.

We’re already seeing this with Siri, Alexa, Google Assistant (and somewhat Cortana), whether on your phone, TV or external connected device. Consumers are beginning to adopt the idea of speaking to technology and speaking to it in a way the devices understand. And what consumers are fundamentally doing is searching. Just instead of the favourite search engine, they’re using search anywhere and everywhere.

This is a big deal. The more these AI assistants develop and begin to comprehend what consumers mean when they speak more fluently in their natural language, the more consumers will begin expecting all search bars/engines/platforms to do the same.

Problem is, most off the shelf search integrated into ecommerce shopping cart providers such as Shopify, Magento or Open Cart have extremely basic search capabilities built in. Same for blogs and pretty much anything else.

How traditional search works

The tried and tested search mechanism has always been a Boolean based operation. All data would be stored on a relational database and use operators such as AND, OR, WHERE, LIKE. So if a customer is searching for a grey, fitted gym t-shirt, the function behind this would look similar to “SELECT * FROM products WHERE title, description, tags LIKE $search”.

This kind of search misses out on the nuances of human language. What it does is look for the searched keywords in the database of products under the title, description and tags. It essentially carries zero meaning (this is an overly simplistic view of something that is a little more complex, but you get the point).

How natural language changes this

Natural language changes the way we can interact with systems. A query like “grey, fitted gym t-shirt” can now be parsed to understand intent & entities.

Intent is what the users query means, in this instance it’s simply a search.

The interesting part comes together with Named Entity Recognition, or NER. These NER algorithms are trained to parse keywords in a given text and extract and assign the keywords to classifiers. So if trained correctly, the query would look something like this in the backend once parsed:

Grey => colour

Fitted => fit_type

Gym => product_attribute

T-Shirt => product_type

So now the system understands the user is looking for a product (t-shirt) which is of a particular colour (grey), is concerned with the fit type (must be fitted), has a use case for this (using in the gym, so that could mean sweat wicking, polyester) which then helps to build a more concise query in the backend that takes into account all of the above.

This natural language search reduces friction in search by understanding the semantic context of the users search phrase or query and can help provide more relevant results. In the ecommerce space, this is the kind of technology every store needs to have as an absolute minimum.

When consumers can search for something as specifically as this and get the relevant search results back, we begin to earn their trust in providing quick, relevant results. Google did the same with web-search which is how they became the web-search authority.

Ecommerce and blog search can benefit from integrating natural language processing AI technology to correctly map out and identify products or articles quickly and concisely.