Sometimes we even get questions about Google Search itself – like, how this whole thing actually works.

有時，甚至會有人來問 Google 搜尋是怎麼運作的。

And while this is a subject entire books have been written about, there's a good chance you're in the market for something a little more concise.

已經有一套書講解了這個原理，但你很可能會想知道精簡版的答案吧？

So, let's say it's getting close to dinner and you want a recipe for lasagna.

假設，晚餐時間快到了，而你想要找...千層麵的食譜。

You've probably seen this before.

你大概看過這個畫面了。

But let's go a little deeper.

但讓我們再深入一些。

Since the beginning, back when the homepage looked like this, Google has been continuously mapping the web - hundreds of billions of pages - to create something called an index.

從最一開始，Google 首頁還是這樣子的時候，Google 就已經不斷地在分類網站，用數以萬計的網頁，來創造「索引」。

Think of it as the giant library we look through whenever you do a search for lasagna or anything else.

就把它想成一個，每當你想要搜尋千層麵或其他東西，就會去的巨型圖書館。

Now, the word lasagna shows up a lot on the web: pages about the history of lasagna; articles by a scientist whose last name happened to be "Lasagna"; stuff other people might be looking for.

千層麵 (lasagna) 這個字出現在許多網站裡，像是有關千層麵歷史的網站、或是剛好姓 Lasagna 的科學家所寫的文章，還有一些其他別人想搜尋的東西。

But, if you're hungry, randomly clicking through millions of links is no fun.

但如果你已經很餓了，點這些無關緊要的連結可一點都不好玩。

This is where Google's ranking algorithms come into play.

這時就是 Google 演算法出場的時候了。

First, they try to understand what you're looking for, so they can be helpful even if you don't know exactly the right words to use, or if your spelling is a little off.

首先，演算法會試著了解你想找什麼，所以他們才能在你不確定用字或拼錯字時，提供最好的協助。

Then they sift through millions of possible matches in the index, and automatically assemble a page that tries to put the most relevant information up top for you to choose from.

然後演算法會過濾索引中數百萬的結果， 依相關性自動組合成一個頁面供你點擊。

Ok, now we have some results.

好，這裡有些搜尋結果。

But how did the algorithms actually decide what made it onto the first page?

那運算法究竟是如何決定哪些網頁可以顯示在第一頁？

There are hundreds of factors that go into ranking search results, so let's talk about a few of them.

有數百種要素決定搜尋結果，那我們來看看其中幾種。

You may already know that pages containing the words you searched for are more likely to end up at the top – no surprise there.

你可能知道，那些含有搜尋關鍵字的網頁，會比較容易在搜尋結果上方，這不意外吧。

But the location of those words, like, in the page's title, or in an image's caption – those are factors too.

但那些字的位置，如在網頁的標題或是在一張圖片的敘述，這些也是要素之一。

There's a lot more to ranking than just words.

要排序的除了字，還有很多。

Back when Google got started, we looked at how pages linked to each other to better understand what pages were about and how important and trustworthy they seemed.

回到 Google 剛成立時，我們探索了網頁之間是如何彼此連結，來更好地了解每個網頁的作用、重要性與可信度。

Today, linking is still an important factor.

今天，連結依然是個關鍵要素。

Another factor is location – where a search happens.

另一個要素是地點，也就是搜尋發生的地方。

Because if you happen to be in Ormea, Italy, you might be looking for information about their annual lasagna festival.

因為如果你剛好在義大利的奧爾梅亞，那你或許想搜尋一年一度的千層麵節日相關資訊。

But if you're in Omaha, Nebraska, you probably aren't.

但如果你是在內布拉斯加州的奧馬哈，那可能就不是。

When a webpage was uploaded is an important factor too. Pages published more recently often have more accurate information, especially in the case of a rapidly-developing news story.

網頁發佈的時間也是個重要元素，較新的網頁通常會有較準確的資訊，尤其是事情會不斷發展的新聞文章。

Of course, not every site on the web is trying to be helpful.

當然，不是每個網站都是有助益的。

Just like with robocalls on your phone or spam in your email, there are a lot of sites that only exist to scam, and every day, scammers upload millions more of them.

如同自動語音電話，或電子郵件裡的垃圾信件，有些網站存在的唯一目的就是詐騙，而這些人每天會上傳數百萬個詐騙網頁。

So just because instantvirusdownload.net lists the words "lasagna recipe" 400 times, that doesn't mean it's going to help you make dinner.

所以就算 instantvirusdownload.net 這個網頁列了「千層麵食譜」400 次，它也沒辦法幫你準備晚餐。

We spend a lot of time trying to stay one step ahead of tricks like these, making sure our algorithms can recognize scam sites and flag them before they make it to your search results page.

我們花了許多時間，試著領先這些詐騙網頁一步，確保演算法能辨識出詐騙網頁並標註它們，這樣它們才不會出現在各位的搜尋結果。

So, let's review: billions of times a day, whenever someone searches for "lasagna" or "resume writing tips" or "how to swaddle a baby" or anything else.

複習一下，每天有數十億次的搜尋，可能是「千層麵」、「寫履歷的秘訣」、「如何用布包好寶寶」之類的。

Google's software locates all the potentially relevant results on the web, removes all the spam, and ranks them based on hundreds of factors like keywords, links, location, and freshness.

Google 的軟體便會定位網路上所有可能的結果，過濾詐騙網頁，然後用數百種要素來排序它們，如關鍵字、連結、地點與時效性。

Ok.

好的。

Good time to take a breath.

做個深呼吸。

This last part is about how we make changes to search, and it's important.

最後，來講講我們是如何改進搜尋功能的，這很重要。

Since 1998, when Google went online, people seem to have found our results pretty helpful.

自 1998 年 Google 上線以來，人們似乎都覺得我們很有幫助。

But, the web is always changing, and people are always searching for new things.

但網路無時無刻不在變化，且人們總是在搜尋新事物。

In fact, one in every seven searches is for something that's never been typed into the search box before, by anyone ever.

事實上，每七個搜尋中就有一個，是從來沒有被任何人輸入過的內容。

So, we're always working on updates to Search – thousands every year.

所以，我們不斷地改良「搜尋」，甚至一年好幾千次。

Which brings up a big question: how do we decide whether a change is making search more helpful?

這又引出下一個重要的問題：我們要如何決定，某個改變是有幫助的改變？

Well, one of the ways we evaluate potential updates to Search is by asking people like you.

這個嘛，其中一個評估「搜尋」更新的辦法，就是詢問像你一樣的人們。

Every day, thousands of search quality raters look at samples of search results side-by-side, then give feedback about the relevance and reliability of the information.

每天有數千名網頁品質評分者會對照各種搜尋結果的樣本，然後就訊息的相關性與可靠性提供反饋。

To make sure those evaluations are consistent, the raters follow a list of Search Quality Evaluator Guidelines.

為了確保這些評估是一致的，評估者會遵守「網頁品質評估守則」。

Think of them as our publicly available guide to what makes a good result... good.

就把他們想成公開可用的守門人，保持搜尋結果的良好度。

Oh, and one last thing to remember: we use responses from raters to evaluate changes, but they don't directly impact how Search results are ranked.

對了，最後一件事，我們會使用評估者的回應來評估，但他們無法直接影響搜尋結果。

So, there you have it: every time you click "search," our algorithms are analyzing the meaning of the words in your search.

總結來說，每次你按「搜尋」，我們的演算法就會分析你輸入的每個字。

Matching them to content on the web, understanding what content is most likely to be helpful and reliable.

將它們與網路上的資訊配對，分析哪個內容可能對你最有幫助且可信。

And then automatically putting it all together in a neatly organized page designed to get you to the info you need.

然後自動把它們整齊地放在同一個網頁，讓你得到所需的資訊。

All in... oh 0.81 seconds.

這全部只要...噢，0.81 秒。

Anyone else ready for dinner?

準備好要吃晚餐了？

Interested in learning more?

想要學到更多？

We've got a whole website dedicated to how Search works.

我們有一整個網站解釋搜尋引擎是如何運作的。

Just click right here.

只要點擊這就好。

Want to read the Search Quality Rater Guidelines for yourself?

想要讀讀網頁品質評估守則？

Click right here.

點擊此處。