A series of improvements to LinkedIn's search system for posts will allow it to deliver faster and more relevant results within a simplified user interface.
LinkedIn explains the process of developing its new system architecture while also addressing the depth of complexities its previous system had. Before, search results for posts were served by two indexes: one for LinkedIn's main feed and one for articles.
It was rather difficult to develop upon due to its complexity, so LinkedIn decided to split the indexes apart. In a new blog post, LinkedIn goes through every step in lengthy detail. Here's a breakdown:
Image credit: Screenshot from engineering.linkedin.com/blog/2022/improving-post-search-at-linkedin, August 2022.
Most of the information in LinkedIn's post was for software engineers making it rather technical, in this post we'll skip past that and provide the details on what the updates will mean for the typical user.
LinkedIn has created a better system to make search results more relevant by focusing on the following aspects:
LinkedIn states that its new system will deliver results from multiple sources and leverage several machine learning techniques to meet the expectations of the user. LinkedIn also utilized human ratings to assess search results and used this data to improve the quality of its new system.
The new LinkedIn system, powered by machine learning, improves upon the previous way of doing things in the following ways:
Relevance: Deeper and real-time signals for member intent, affinities, and interests enables more personalization
Diversity: Discovery of potentially viral content is increased for trending queries while reducing the rank of similar duplicated content.
Ranking: Post-related metadata from the index is used to improve the ranking of posts when mixed with other types of results.
Navigation: There is a brand new user interface allowing users to search for posts from specific authors, posts viewed recently, match quoted queries and more.
There has been a 20% increase in positive feedback based on results delivered by its new system. LinkedIn refers to results that are highly relevant to a user's search as "pertinent results", these results have led to a click-through rate improvement of over 10%.
By providing more relevant and diverse posts based on the searcher's geographic location, their social network, and preferred language LinkedIn has reported a 20% increase in messaging within the network of the searcher. The overall time for search results to be delivered to users has decreased as well.
Here are a few future improvements LinkedIn plans to make:
Click here to take a look at LinkedIn’s full blog post for all of the details behind these changes.
Source: LinkedIn
Image credit: Later.com
