
Details of Search Quality
The NY Times has published a huge article on how, who and why at Google is changing ranking algorithms. The article provides many interesting facts about how the Search Quality department works. The sent journalist spent the whole day there.
As you know, fine tuning of the results occurs in the search engine almost every day. Google’s search-quality department makes about six major and minor changes to the ranking algorithms per week. The process is led by 39-year-old Indian-born programmer Amit Singhal. Like other elite programmers, he was awarded the honorary position of Google Fellow.
After joining Google in 2000, Mr. Sinhal has developed a sophisticated document ranking system that takes into account more than 200 factors that are called “signals” in professional slang. PageRank is just one of those signals.
Each of the 10,000 Google employees can use the special Buganizer system to report a particular search engine error. All these messages (about a hundred a day) are reviewed by Amir Singal and several other top Google executives, including Matt Cutts , the head of Google’s anti-spammer department.
Some problems are quite obvious and can be fixed without thinking. For example, recently a search for the phrase [French Revolution] yielded more results about the last presidential election in France than about the real French revolution of the times of Louis XVI.
Other problems are more complicated. So, in 2005, the problem was considered why there is no local store in the issuance of [teak patio Palo Alto] upon request, which is 100% semantically identical to the request. This is the Teak Patio store in Palo Alto (Theteakpatio.com website). Understanding this problem, Amir Singal launched “one of the most secret and closed internal programs” called Debug to analyze the results of a search for a specific search query. It became clear that there was no store in the store because incoming links were not enough weight, because small regional sites usually have low PR. To correct this and similar situations, Amir Singal made adjustments to the ranking algorithms. These changes affected all the smallest regional online stores in the world.
In the second half of last year, many complaints in the Buganizer were related to the "freshness" parameter. This is one of the parameters that affects the output. Previously, those pages that have not been updated for a long time were considered more authoritative. However, after the launch of the Google Finance service, Mr. Singal found that he could not find this site on the first page of search results. It became clear that the algorithms needed to be changed and give “fresh” pages more weight.
Amir Singal couldn’t make such a global change on his own, so this spring he presented his calculations at the weekly meeting of leading Google search specialists. He showed a presentation on the initiative, which was called QDF (query deserves freshness). The main idea of QDF is to find those pages that are mentioned in blogs and forums and, accordingly, are considered “hot”. Colleagues decided that QDF deserves a two-week test in one of the data centers.
This is how Google’s ranking algorithms are set up.
via NY Times
As you know, fine tuning of the results occurs in the search engine almost every day. Google’s search-quality department makes about six major and minor changes to the ranking algorithms per week. The process is led by 39-year-old Indian-born programmer Amit Singhal. Like other elite programmers, he was awarded the honorary position of Google Fellow.
After joining Google in 2000, Mr. Sinhal has developed a sophisticated document ranking system that takes into account more than 200 factors that are called “signals” in professional slang. PageRank is just one of those signals.
Each of the 10,000 Google employees can use the special Buganizer system to report a particular search engine error. All these messages (about a hundred a day) are reviewed by Amir Singal and several other top Google executives, including Matt Cutts , the head of Google’s anti-spammer department.
Some problems are quite obvious and can be fixed without thinking. For example, recently a search for the phrase [French Revolution] yielded more results about the last presidential election in France than about the real French revolution of the times of Louis XVI.
Other problems are more complicated. So, in 2005, the problem was considered why there is no local store in the issuance of [teak patio Palo Alto] upon request, which is 100% semantically identical to the request. This is the Teak Patio store in Palo Alto (Theteakpatio.com website). Understanding this problem, Amir Singal launched “one of the most secret and closed internal programs” called Debug to analyze the results of a search for a specific search query. It became clear that there was no store in the store because incoming links were not enough weight, because small regional sites usually have low PR. To correct this and similar situations, Amir Singal made adjustments to the ranking algorithms. These changes affected all the smallest regional online stores in the world.
In the second half of last year, many complaints in the Buganizer were related to the "freshness" parameter. This is one of the parameters that affects the output. Previously, those pages that have not been updated for a long time were considered more authoritative. However, after the launch of the Google Finance service, Mr. Singal found that he could not find this site on the first page of search results. It became clear that the algorithms needed to be changed and give “fresh” pages more weight.
Amir Singal couldn’t make such a global change on his own, so this spring he presented his calculations at the weekly meeting of leading Google search specialists. He showed a presentation on the initiative, which was called QDF (query deserves freshness). The main idea of QDF is to find those pages that are mentioned in blogs and forums and, accordingly, are considered “hot”. Colleagues decided that QDF deserves a two-week test in one of the data centers.
This is how Google’s ranking algorithms are set up.
via NY Times