Twitter now doesn’t want to show you what is new in its search results but instead wants to show who tweeted in the best way. Now the social media platform has moved further away in search from reverse chronological order toward relevance order. You will now be shown relevance-ordered tweets across the platform, first on the results page whenever you look for something using the search bar, notes Mashable.
Search results by relevance instead of time
According to a Twitter spokesperson, the change was made in September, but it was announced publicly on Monday. The “latest” filter is still there if you want to see the most recent live tweets. Also the photos, videos and news filters are there as well.
Earlier this year, the changes in Twitter’s timeline algorithm caused the internet to briefly lose its mind, but changes to search are unlikely to cause even a stir. But both changes show the social networking site’s obsession with the hard concept of “relevance,” says Mashable.
Twitter’s opt-in timeline changes, which were announced in February, focused primarily on what it assessed to be “the best tweets” in the user’s timeline instead of simply showing users the most recent tweets from the people they follow. It looks like the same type of programming that the micro-blogging site uses on its main timeline is going to implemented with the search results as well. However, it will come with some variation because it has to parse more data, as it is pulling Twitter’s entire database and not just the people the users follow, notes VentureBeat.
Relevancy not simple on Twitter
Twitter claims that based on early trials, there has been more engagement with tweets and search results with more time spent using its network. The most recent tweets in search do not necessarily contain the most useful information, but relevancy is complex.
In a blog post, Lisa Huang, a senior software engineer on Twitter’s search quality team, explained how hard it is to prioritize so-called “relevant” tweets in the search results. The team is using machine learning to assist in deciding how tweets will be ordered.
“A person’s behavior on Twitter provides an invaluable source of relevance information,” Huang stated. “Using this information, we can train machine learning models that predict how likely a Tweet is to be engaged with (Retweets, likes and replies). We can then use these models as scoring functions for ranking by treating the probability of engagement as a surrogate for the relevance of Tweets,” she said.
Huang also wrote that users could be searching for popular tweets to better understand the context around the search query.
