Why AI-powered query categorisation is a smart choice

Users search. We all look for things across the expanses of the Internet, throughout the intranets and proprietary data channels that we are granted access to, across the expanses of the e-commerce landscape, inside our favourite applications for content and around an almost infinite variety of other digital touch-spaces from airport kiosk computers to electronic soda vending machines.

While the drinks options in even the smartest new-age drinks robots are limited enough to be searched via a basic user interface - and the user’s intent is clearly ‘I’m thirsty, I need a drink’ - the nuances of human search and their underlying intent and purpose are generally far more convoluted and complex. If we’re not just offering options for Coke and Fanta (other sparkling beverages are also available), there’s a clear need to engineer some product or service intelligence into the way any given search function operates.

Specialist end-to-end AI search and discovery platform company Algolia works in the business of search as a facilitating function inside what are often expansive information repositories; distinguishing ‘search’ as an action performed inside a platform, web service, application or other digital entity as opposed to web search as in Google.

Now looking to provide a search and discovery function that automatically associates a user’s query with an appropriate product or service category (or categories), the company has this year spoken about its AI-powered Query Categorization offering. This is a software tool designed to enable merchandisers, business users and non-tech practitioners in e-commerce (or media companies, where content search is similarly large) to understand their consumers’ intent as they begin a search for items.

Algolia AI search: How Query Categorization works

In terms of how this function works, Algolia explains Query Categorization (capitalised as a branded offering) as a means of enabling organisations to ‘operationalise their strategies’ across entire product categories more efficiently. In a world where product and service search is a) often done in nuanced idiomatic-rich natural language and b) done so via speech search, there is a need to channel user search requests into appropriate categories so that companies can organise, prioritise and operationalise their supply chains and delivery mechanisms in the most efficient manner. Here that’s done at the back end so that in a few clicks, the business can accurately predict the correct category associated with search queries, which clearly helps create time-saving cost reductions.

Using customer interaction data and advanced AI algorithms, this form of query categorization can automatically predict the true intent behind users’ queries and connect to the correct categories of products. The AI model is trained by each product catalogue (with a full view of what’s inside it, obviously) and its users, producing the best results for each specific business. This reduces the setup and ongoing maintenance of online product catalogues by developers (that’s web developers and software application development developers) as well as product managers and non-tech practitioners who get involved with online sales channel processes such as marketing managers and domain specialists.

An end to ‘null results’

Coupled with its Algolia NeuralSearch technology, the company says that Query Categorization offers tremendous value and increased precision (accuracy) and enhanced recall (completeness) of search results. It’s all supported by autonomous prediction of the most relevant product categories. What it represents (and we have all experienced search functions that deliver a zero) is what Algolia promises as a route to virtually eliminating ‘null results’ for consumers. 

“Query Categorization represents our commitment to continuous innovation and pushing the boundaries of what’s possible. Using AI, [this technology] automatically and accurately predicts the product categories associated with a consumer’s specific query and then connects to the right products based on the online business catalogue or content, whether it is an e-commerce site, movie app, or other content collection,” said Bharat Guruprakash, Chief Product Officer, Algolia.

As a working example, let’s say a query for ‘milk chocolate’ will be connected to a category comprising chocolates, whilst ‘chocolate milk’ will be connected to a category of dairy products and a query for ‘Charlie and the Chocolate Factory’ will be connected to a movie category. Understanding the difference between all three is crucial in our current digital age.

One UK online grocery chain gained upwards of 15% revenue boost after only two months when using Query Categorization in production. Additionally, a Swiss department store chain experienced a 22% increase in their CVR (conversion rate) from 4.23% to 5.17% after two weeks of use.

A significant equalizer

Guruprakash describes this tool as a significant equalizer for many companies that compete with large, established industry players, like ‘big box’ retailers, without large teams of data scientists and AI experts. In many instances, these organisations attempt to ‘do it alone’ and imitate big players by building highly manual processes to sort search results and match them with ranked products and categories. 

Algolia is assertive on this point and says that Query Categorization solves this problem in only a few clicks and empowers these organizations to dynamically map queries to aligned categories and, importantly, at enterprise scale. Presented as a dedicated section in the Algolia UI (user interface), users can set up AI models and explore predictions in an intuitive fashion. 

Because every technology worth its salt is also presented via routes which democratise its functions to non-technical users, Algolia has also provided the ability to perform automatic filtering and boosting on predicted categories using a no-code environment that business users and non-tech practitioners can control directly to increase the relevance of their user’s results.

As noted, this is not web search, Google has that covered despite the best efforts of Bing, Baidu, DuckDuckGo and dear old Yahoo! This is search as an integral part of the way we now build data services to serve user requests in applications spanning every industry vertical.

See also: Taking generative AI to production: What CTOs need to consider