Ten recent Google algorithm tweaks

16 November 2011

In August Google published a post on their Inside Search blog which gave us a behind-the-scenes look at the process involved in their algorithm changes.

The post includes a video in which Rajan Patel, a Google 'search scientist,' explains that Google made roughly 500 changes to the algorithm in the previous year.

Ranking engineers suggest ideas for improving the search results, the algorithm change is then manually tested by 'raters.' The raters are presented with two side-by-side search results and asked to rate which one is better. The change is then tested live in a sandbox. Google divert a random selection of users to the sandbox which is running an algorithm with the new changes. From the data gathered, Google evaluates the success of the suggested changes. If they show a significant improvement to the user's experience they can be rolled out into the live environment.

This month Google have posted a follow-up which talks in more depth about 10 of the 500 improvements made yearly.

Translating search results on-the-fly

When there is little information available in a language such as Catalan or Welsh, Google translates relevant English pages and displays the translated titles directly below the English titles in the search results.

More relevant snippets

Snippets are now much more likely to be taken from the content of the page rather than the description tag or wording contained in menus and navigation.

Better page titles

Google says they are "de-duplicating boilerplate anchors" which means they are taking less notice of the link text of links which are pointing to the page when they build the page title. The result should be more relevant titles especially where there is no title tag on the page.

Length-based autocomplete in Russian

The number of long query predictions has been reduced in Russian language queries and falls in line with what we currently see in English.

Extending application rich snippets

This week Google announced rich snippets for applications which enables people who are searching for software to see details, such as price and reviews by users. These now appear within the search results. The change extends the coverage of application rich snippets, so they will be available more often.

Retiring a signal in Image search

Google have decided to retire a signal which was previously introduced in the image search algorithm which related to images that had references from multiple documents on the web.

Fresher results

At the beginning of November Google announced, that they've made significant improvements to how they rank fresh content. Using their new Caffeine web indexing system, which allows Google to crawl and index the web for fresh content quickly and on an enormous scale, Google is now introducing this fresh content where appropriate affecting roughly 35 percent of total searches.

Refining official page detection

With this change, Google adjusted how they attempt to determine which pages are official. This will tend to rank official websites even higher in the search rankings.

Improvements to date-restricted queries

This affects the way Google handles result freshness for queries where a user has chosen a specific date range, helping to ensure that users get the results that are most relevant for the date range specified.

Prediction fix for non-latin character queries

This change improves how Autocomplete handles IME queries. Autocomplete was previously storing the intermediate keystrokes needed to type each character, which would sometimes result in gibberish predictions for Hebrew, Russian and Arabic.