An insider's look at Bing's search relevance system

Part of Bing's appeal, Microsoft would have us believe, is that the results it returns are more relevant than the competition. Regardless of whether you feel like they're suffering or failing in that area, it's up to Bing's 'Human Relevance System,’ or the HRS, to make sure that the results you get are as accurate as possible.

Believe it or not, there really are people out there who make their living by looking at search results all day. Until recently, Microsoft has kept its search quality guidelines close to its chest, but that was before a former 'judge,' as the folks who rate search quality are known, revealed Bing's quality guidelines to Search Engine Land.

Bing’s relevance rating system for landing pages, or ‘LPs,’ is actually pretty in depth – the HRS guidelines document is 52 pages long – but in a nutshell, it comes down to four simple, direct questions:

1. Intent: Does the LP content address a possible intent for the query?
2. Scope: Does the range and depth of the LP content match what the user wants?
3. Authority: Is the trustworthiness of the content on the LP appropriate to the expectations of the user?
4. Quality: Do the appearance and organization of the LP providing a satisfying experience?

From there, the judges rate each result on a scale from ‘strongly’ to ‘poorly.’ From there, the judges use something called a Rating Matrix, which you can check out below, to decide how to rate the results.

The final five ratings range from ‘bad’ to ‘perfect,’ and take a variety of situtations and intents into account, as well as the results themselves. For example, the guidelines say that generic entries never have a ‘perfect’ landing page, since there are so many variables involved, but clearer searches, for ‘Bob Dylan,’ for example, should have an official, ‘perfect’ page at the top of their results.

On the other hand, a ‘bad’ rating is reserved for spam and sites that contain malware. According to the Rating Matrix, a bad rating only addresses 1% of the searcher’s intentions. Beyond that, there are a couple more non-standard ratings, like ‘detrimental’ (really, really bad, particularly, the guidelines emphasize, if it shoves porno down the throats of users who weren't looking for it – pun intended), to ‘no judgment,’ for sites that can’t be accessed for one reason or another.

There’s one more factor that Bing’s HRS has to take into account: freshness. As the name implies, a freshness rating attempts to place more recent (and therefore more relevant) results at the top of the page. At the same time, it takes into account that some searches don’t want fresh results, and ideally distinguishes between those. A search involving a certain year, for example, doesn’t need to be fresh.

All in all, Bing’s HRS system isn’t too different from what other search engines use. Of course, that doesn’t necessarily mean that it’s not as good as its competitors – if the results are more accurate to begin with, HRS guidelines don’t really matter. And besides, how much room for improvement is there when it comes to something that's so narrowly right or wrong? It's a little hard to go beyond that, isn't it? It just goes to show that it takes old fashioned human judgment to make sure that everything is running smoothly, no matter how good computers get.

Source: Search Engine Land | Image via Search Engine Land

Report a problem with article
Next Article

RIP ChevronWP7, thanks for all the memories.

Previous Article

Skype for Metro leaks, we go hands on

7 Comments - Add comment