Recent Google Patent Incorporating Social Signals Into Search Results
It is no doubt that web structure graphs are much different from social graphs. For example, the Strong Triadic Closure Property does not necessarily apply to web structures, but is very relevant in social graphs. When discussing web search methods and algorithms, we took unique approaches to ranking results, which didn’t relate to social graphs. Google uses the PageRank algorithm to rank its searches. However, social graphs may play a bigger role in newer web search algorithms.
Dave Davies analyzes a recently approved Google patent that incorporates “social signals” into its web search algorithm. When Davies uses the term “social signal”, he is referring to components in one’s social graph. According to the patent, the social graph is not constrained to just social connections via Facebook, Twitter, etc., but “… all things that could be considered social. This includes email, chat programs, blog posts, reviews, and so on.” Thus, the social graph extends beyond just friendship graphs like we had focused on during class. Davies explains the essence of the patent, which is to create a set of search results that have been modified based on one’s social graph (which will inherently be different for each person). He goes on later to say that other characteristics from the social graph, such as degrees of separation between people, may also affect the search results. The specific implementation of these kinds of search results is still unclear. It could be that search results calculated using these “social signals” are clearly differentiated between regular search results, but they could also be combined. Nonetheless, the main goal is to return relevant search results that would have been ranked lower with something like PageRank. Davies notes that in the patent, Google understands that one’s social graph may have to be large enough to hold some significance.
This new method of ranking web searches brings up many different questions as to how this would affect the relevance of returned web searches. It could be the case that when someone make a search, they’ll want similar information as the people they interact with. This could be extremely useful in the case where people who are close to each other want to stay updated with one another. However, what about when someone wants to discover new things that those around them are not necessarily concerned with? In class we had discussed the “strength of weak ties”. We discussed that most opportunities and new information actually came from people that we didn’t have much interaction with. Thus, would it be the case that incorporating “social signals” would decrease the relevance of search results? Or perhaps the algorithm would be able to recognize such issues and actually use weak ties instead.
There are many ways in which using social graphs in search ranking algorithms can be useful, but the true effects of such methodology are still unclear. Davies explains how Google has attempted to incorporate similar improvements before, with this patent being another attempt. However, he notes that this method is just a patent, and is not entirely indicative of the future of Google’s web search algorithm. Regardless, it is interesting to see how new developments can be made in order to improve the relevance of what we search based on each individual’s social interacts as opposed to one generic page ranking for everyone.
Source: http://searchengineland.com/google-patent-using-social-signals-impact-rankings-259656