Coveo Machine Learning Features
The Coveo Machine Learning (Coveo ML) service currently offers the following features:
In short, the ART feature learns what search users seek and delivers it.
In more detail, ART analyzes user behavior patterns from many Coveo Usage Analytics (Coveo UA) search visit actions (e.g., query reformulation, clicked results1, whether a support case was submitted) to understand which clicked results and content lead to successful outcomes, such as self-service success. It automatically adjusts future search results, so that the best-performing content always rises to the top.
1: ART models also learn from actions performed by users within a given search result (e.g., clicking a search result Quick View or attaching a result to a Case in a Coveo Insight Panel).
ART excels with popular and ambiguous queries, in which users only enter one or two terms, as well as with paragraph-sized queries expressing long descriptions. ART can handle common typographical errors, and it learns implicit synonyms. When your Coveo index includes content from sources that index permissions, ART queries the index to ensure that it only recommends items that the user performing the query is allowed to access.
In practice, ART boosts the ranking weight of recommended items so that they appear among the top search results. Members with the required privileges can configure and activate ART in just a few clicks.
By default, ART model recommendations are based on the language of the user’s query as well as the search interface in which the query is performed. ART models build a submodel for each language, and then apply filters on these submodels for each search hub and search tab to better tailor the provided recommendations to the user’s context.
Items are boosted only if they were clicked in the same search interface as the current query.
The QS feature recommends significantly more relevant queries to users as they type in the search box.
The Coveo ML QS feature:
Identifies exact, partial, or fuzzy matches with typed characters anywhere in any individual keyword appearing in any order.
Stems query suggestion keywords to remove duplicates.
Offers the most relevant recommendations by ranking query suggestions, considering:
Only considers queries that were followed by clicked search results a specific number of times (see Reviewing Coveo Machine Learning Query Suggestion Candidates). This prevents infrequent queries from polluting the suggestions.
Members with the required privileges can configure and activate Coveo ML QS in a few clicks. Developers can leverage the feature in the desired search interface.
The ER feature learns from your website user page and search navigation history to return the most likely relevant content for each user in his current session. The results that are returned by the Coveo ML ER service can be included in a search page or in any other web page, such as in a side panel window.
The recommendations can be interpreted as “
People who viewed this page also viewed the following pages”.
Your company offers product technical documentation and Q&A content on several public websites for customer end users, administrators, and developers. These websites are configured to send all views to Coveo UA. Your website pages include Recommended Articles side panel windows.
The recommendation algorithm is based on the co-occurrence of events such as view events within a user session. When two events co-occur with unusual frequency within sessions, the algorithm learns that they’re linked. When one event is seen, the model recommends the other.
ER model suggestions are provided based only on the user language, since view events aren’t logged from a search hub.
The DNE feature learns from usage analytics events to pertinently order facets and facet values according to user queries. More precisely, Coveo ML DNE models analyze queries and target specific user behaviors (e.g., clicked results, facet selections) to make the most relevant facets appear at the top for a given query.
Coveo ML DNE models also reorder facet values within a given facet to make the most popular values appear at the top. To do so, the models use the search events performed by previous users, who have selected certain facet values for a specific query.
The Coveo ML DNE feature uses its facet value ranking to boost search results. The model uses the most popular facet values for a certain query and applies query ranking expressions (QREs) to boost the search results whose field values match the values of those facets.
You enable a Coveo ML DNE model. When your search interface sends a query to the Search API to request facets, the DNE model can modify that query in the query pipeline. It applies insights gained from the analysis of past customer behavior and determines that users are most likely to sort search results using the Brand and Price facets. Your search page now displays facets in the following order:
Before you enabled Coveo ML DNE, the Brand facet displayed its facet values in the following order when customers searched for
Since the Coveo ML DNE model determined that customers are more likely to shop for Apple smartphones, the model modifies the user query to boost Apple smartphone result list items.
1: You must contact Coveo Support to take advantage of the DNE autoselection feature.
The DNE autoselection feature automatically selects facet values according to the end user query. The feature learns from your end users behaviors to understand which categories are the most relevant according to their current browsing task.
In a Coveo-powered commerce interface, an end user searches for queen sheets. Based on the current context and recorded usage analytics data, the DNE model automatically selects the
bedding value from the Product category hierarchical facet.1 This filters out potentially irrelevant products such as queen mattresses, sheets of paper, or albums by the band Queen.
1: You should consider configuring the DynamicHierarchicalFacet component in search interfaces that leverage the DNE autoselection feature.
The Coveo ML PR feature takes advantage of Coveo UA to suggest relevant products to end users based on their past and present interactions with your Coveo-powered commerce implementation.
Coveo ML PR enhances your customers’ shopping experience by offering them products that suit their profile, context, and buying behaviors. In order to provide relevant suggestions, the model continuously learns from your end users’ feedback by scoping their buyer profile and analyzing their positive and negative interactions with different products. Thanks to its multiple algorithms, Coveo ML PR can easily adapt its approach to your digital commerce strategy.
To deploy Coveo ML PR in a Coveo for Commerce solution, see Leveraging Machine Learning Product Recommendations.
Depending on your context, you can leverage one or more of the available PR strategies:
When leveraging the Interest-based recommender strategy, the model suggests products to the current user based on their general interests. To achieve this, the model learns from users’ previous actions, and uses this information to find other customers that share similar browsing patterns. The model then suggests products that have been previously browsed by customers who share similar interests with the current user.
Frequently Bought Together
When leveraging the Frequently bought together strategy, the model suggests products frequently bought with the current product based on the shopping cart of other users. In other words, the model analyzes the items customers often buy together with the product that the user is currently looking at.
As a user of an electronics retailer commerce website, you are currently viewing a gaming computer. Based on other customers’ purchase history, the model suggests gaming mice and keyboards that were bought together with the computer you are currently viewing.
Frequently Viewed Together
When leveraging the Frequently viewed together strategy, the model suggests items related to each of the current products based on other users’ interactions with your commerce implementation in a single visit. This strategy thus analyzes users’ navigation behavior to provide the user with a selection of similar products that are often seen together in a single shopping session.
As a user of an electronic retailer commerce website, you are currently viewing the ABC gaming computer. Based on the interactions made by other users that have browsed the same computer in their shopping session, the model suggests products that have frequently been seen together with that computer in the same shopping session.
Since other users have also browsed the 123 and DEF computers along with the ABC gaming computer that you are currently viewing, the model suggests these computers to you.
When leveraging the Cart recommender strategy, the model analyzes frequent buying patterns by grouping sets of products that are frequently purchased together in the same shopping cart.
Given a cart, it will recommend products that were frequently bought together in previous similar carts. This strategy is ideal for cross-selling items on a shopping cart interface.
As a user of an electronics retailer commerce website, you are currently viewing your cart where you have a TV and a TV stand. Based on other customers’ purchase history, the model suggests TV accessories, such as HDMI cables, that were bought together with the same TV and TV stand you have in your cart.
Popular Items (Viewed)
When leveraging the Popular items (viewed) strategy, the model recommends the most viewed products. This strategy is useful when you want to display recommendations on a landing page and also used as a backup solution for other strategies when they didn’t gather enough data to provide relevant recommendations.
Popular Items (Bought)
When leveraging the Popular items (bought) strategy, the model recommends the most purchased products. This strategy is useful when you want to display recommendations on a landing page and also used as a backup solution for other strategies when they didn’t gather enough data to provide relevant recommendations.
|Action||Service - Domain||Required access level|
Machine Learning - Models
Search - Query pipelines
Machine Learning - Models
Search - Query pipelines