LinkedIn Outlines Improved Semantic Search Capacity

LinkedIn Outlines Improved Semantic Search Capacity

Earlier this year, we explored the role of hashtags on LinkedIn and questioned whether they’re still necessary for your posts.

LinkedIn’s history with hashtags has been inconsistent—initially resisting them, later embracing them, then removing them, only to reintroduce them in 2018 with a greater focus on discovery.

However, since then, LinkedIn has sent mixed signals regarding the importance of hashtags, even removing them from the Creator Mode display on profiles. So, we reached out to LinkedIn for clarity. Their response was that while hashtags can aid in discovery, their algorithms now consider a wider range of contextual factors when matching search queries.

Recently, LinkedIn provided more insight into this, offering an engineering overview of its Search system’s evolution, highlighting the integration of semantic matching, which uses a broader set of elements to interpret search prompts.

According to LinkedIn:

“Upon reviewing our capabilities, we noticed room for improvement in our content search results for complex queries. At times, we either returned no posts because none contained all the query’s keywords, or we returned posts with the keywords but without adequately answering the question due to a lack of conceptual understanding. However, our analysis indicated that our search index often contained posts that could correctly answer the query, even if they didn’t include all the keywords. This led us to introduce semantic matching in our content search engine.”

The overview details how LinkedIn now incorporates more elements to better understand and respond to search queries, considering the entire query text (including concepts, not just keywords) and personalization based on past user interactions.

LinkedIn evaluates the success of these updates through topic matching and dwell time, reflecting the relevance of search results.

The results have significantly improved overall discovery:

“Our new content search engine with semantic capabilities has enabled us to answer complex queries like ‘how to ask for a raise?’ and has improved on-topic accuracy and long-dwell rates by over 10%. We also saw a positive impact on LinkedIn’s overall sessions, as users are more engaged with the platform when they receive better search results.”

So, does this mean that hashtags are no longer necessary, given that LinkedIn’s discovery system is now more adept at matching queries beyond just keywords?