Search engines look at Relevance and its dynamic nature in ranking web pages. But there’s another element that’s important to consider when looking at how pages are ranked and ordered.
Materiality and Evidence
When evidence is introduced into a civil or criminal case, a judge not only weighs how relevant that evidence might be before it’s turned over to a jury, or before he or she considers it. The judge also decides how material that evidence might be.
For example, a defense attorney might try to introduce good character reference testimony in a case while defending a 30 year-old accused of murder. The testimony is relevant because it’s about the defendent. It goes to show something about him that a finder of facts might find useful. But, what if the person being offered to testify is the defendant’s kindergarden teacher? What if the teachers testimony is about how the defendant would always take his naps on time, played well with the other kids, and never showed signs of anger. While relevant, the testimony just really isn’t material. As testimony about the defendent from around 25 years earlier, it just isn’t that important.
A judge may also look at other factors in deciding whether evidence is material or important enough to enter into a case. One of these is called judicial economy. As an example, let’s say that our accused murderer from the last paragraph committed his crime in the middle of the infield at Yankee stadium in front of a crowd of 50,000 baseball fans. Almost everyone in attendence witnessed the crime. But you won’t see all 50,000 being summons to appear in court to testify. Their testimony is relevant, and material, but it doesn’t serve the best interests of justice to have that many people testifying if they are all going to give substantially the same testimony.
Another factor that a judge may consider while determining how important evidence might be is prejudice. A prosecutor wants to bring in evidence that our murder defendant was driving 10 miles over the speed limit and was given a citation, a week before the murder. This fact involves the defendant, and it’s a criminal act, but it really doesn’t show one way or another that the defendant might have committed the murder, and it might be prejudicial to have that brought before a jury. To have a jury think that the defendant is someone with a complete disregard for the law might prejudice them when it comes to deciding upon just the facts involving the murder itself.
Evidence in legal cases needs to be both relevant and material, and it needs to pass certain thresholds determined by a judge before it can be brought into a legal proceeding.
Materiality and Search Engines
Search engines also attempt to return results in response to queries that are both relevant and material. Sometimes those are referred to as query-dependent and query-independent signals. A query dependent signal includes some information about a query itself in determining a ranking score for a particular page. A query independent signal doesn’t take any specific query into account to determine a score for a page. The two types of scores can be combined to acheive a final ranking score.
An example of a query independent ranking signal is PageRank, which looks at the quantity and quality of links pointing to a page to determine how important that page might be. While some pages might have a very high PageRank score, like the home page of Google, that score won’t help them rank for queries that they just aren’t very relevant for.
Freshness is another area where materiality plays a role in determining scores for pages. Search Google for [microsoft windows] and you won’t see pages about Windows 3.1, Windows 95, Windows XP, or other older operating systems. Among the results you’ll see some generic pages about the topic, as well as a few pages about Windows 8. Like the character reference testimony from our kindergarden teacher in the murder case example, those results just aren’t very material.
A number of the patents from Google and Microsoft describe features of web pages that might go into a static ranking or a quality score for a page. These may involve a very wide range of features that could be found on a page, from the number of words, to a reading level, to page layout and even the locations of ads on a page, whether there’s a privacy policy, if there’s poor grammar, and others. They could also involve looking at things like how frequently a page acquires new links, or how frequently the content of a page changes over time. A Microsoft paper published last year referred to how credibility of a page might be used to re-rank search results as well.
Take Aways
The rankings of pages and other items (videos, images, news, books, music, maps) in search results may be based in part by how relevant they might be to a query as well as how relevance is being defined based upon the query used. Those rankings also can depend upon how material a result might be in combination with that relevance.
There are some types of ranking signals that combine both relevance and materiality such as a personalized PageRank, or a social reputation rank that might assign people reputation scores based upon topics that they contribute to social networks and interact with others on, and those reputation scores can play a role in the rankings of pages in social search results, and possibly in Web search results in the future as well.