badClicks and goodClicks: mainly refer to the click behavior of users in search results. are clicked, it is considered as a plus point, otherwise it is a minus point.
lastLongestClicks and lastGoodClicks: mainly measure whether clicks decay over time, taking into account timeliness;
unsquashedClicks, unsquashedImpressions and unsquashedLastLongestClicks: Mainly to standardize the data to ensure that no extreme data affects the overall value.
There is another page that is even more core. Google directly said, " We don't understand the content of the webpage, we just pretend to understand it." Google is more about observing people's reactions to the denmark mobile database webpage, and then remembering their reactions to understand whether they are satisfied with the results. If you think about it carefully, it is indeed true. Google's frequency calculation and vector calculation of words, natural language processing, and deep learning are all calculations of language probability, and cannot understand the true semantics AI does not understand it either.
The antitrust materials also mention the ranking process, which is a two-way dialogue with users. First, give users some initial rankings based on content and external links, then observe user interactions and adjust rankings. Therefore, user interaction data is very critical . The user interaction data here is not just dwell time and bounce rate, but a series of indicators such as browsing, clicking, and clicking on other search results to reflect user satisfaction with search results. It can also be logically deduced that Google's ultimate goal is to meet users' search needs, and user interaction is directly related to search satisfaction.