By Guest Author Keren Dagan

In the this post and the next one I will introduce building and using social graphs in different search engines. In this part I will introduce the social graph concept and will go over some of the existing applications that leverage the graph structure. In the second part I will go over alternative graph structures and what kind of problems they could applied to solve.

Introduction: Social graphs

I use the term ’social’ as it is used on the internet: bringing people together to interact and share things. There are many social networks out there where people are sharing media (images, audio and video), sharing domain knowledge and expertise, sharing and playing games, sharing life streams. In some cases they join the network(the object) in order to network (the verb). In most cases there is a social object in the core of the system that people interact and share around it (inspired by Jyri Engeström from Jaiku). Social networks empower us to express ourselves and to build relationships with others on the web providing multiple means of communication, sharing and interaction. Examples:  flickr, facebook, YouTube

When we join a social networks we add another node to the existing social structure. As we start interacting with other people on this network we make additional links.  People, connections and interactions create the social graph. There are multiple ways to create social graphs and I will elaborate about them more later. Capturing and leveraging this graph can make the network even more valuable in multiple ways.

*  The social graph can help members to find other members and make more connections
*  The social graph can help to highlight popular members or objects
*  The social graph can help to make money  - more targeted advertising
*  The social graph can drive crawling and indexing of content more efficiently and keep the network clean from spam

Monitoring and looking at a graph’s patterns can help the website to learn about its online community and improve its offering.  A graph is mainly nodes and links. In most social networks the nodes are people and the links indicate relationships such as friends, fans or followers.

Social search engines:
There are growing types of social search engines.  Some are used for finding content, some for finding people and some for both.  There are two ways to discover content. The first is through a user’s submission and the second is through crawling. The type that focuses on content and users’ submissions usually don’t emphasize the use of the social graph. Examples: del.icio.us and ma.gnolia. They do provide the option to add friends, create groups or bundles, but you can’t learn much about the relationships between members.  Social search engines for content are based on the assumption that my friends from the graph have the best knowledge and the entire graph wisdom will bring good content to the top. These two examples, also know as social bookmarking, offer content that was discovered, bookmarked and tagged by the service members only. If you have a doubt about the usefulness of these search methods, just try typing the term free service, or free software on Google. Social search engines rely on three key features to succeed as I wrote here:

* the search - the basic search engine and the members’ participation in tagging the content (organizing)
* the graph growth - through the option to grow the network both from the outside and the number of links within
* the content growth  - through the option to import your content from anywhere the service has access to

Social graphs search engines:
The search engine that uses a graph to drive its content discovery (i.e crawling and indexing) may also leverage the graph for finding new relationships. I can think of two examples that are fairly similar: Delver and Nsyght: These search engines work based on the assumption that the collective knowledge of my friends, gathered from multiple social networks, will bring better content to the top. Delver provide the options to look for people too. Nsyght emphasizes content. When you join their service you provide them with your personal access information to your multiple social network accounts and grant them the permissions to import your contacts, media and bookmarks from the other web sites. Delver is using this graph to find people and relationships too.

Another case of a search engine using graphs to find cleaner content is Twingly the blog search engine. Twingly is working to provide spam free search results using a graph of known and approved bloggers (a white list). They expend their crawl away from these bloggers with the assumption that good bloggers will not link to spammers. Sometimes when this assumption proves wrong and the crawler falls on a cluster of poor resources, it is easy to identify the source of the problem and then cut off an entire branch.

The search engines that use social graph to find people and relationships usually bring the graph to the front. The most notable ones are Geni and LinkedIn. Geni lets you build your family tree and LinkedIn to build your professional network. LinkedIn allows you to search for people, organizations and relationship. In LinkedIn when you find the person or organization you’ve been looking for, you are also able to see how you might be connected, as well as how many degrees of separation between you, and who can help you connect. The site also use the graph to tell you about people that you might know and are not yet in your network. LinkedIn is a good example how the site is trying new things to increase interactions between nodes on its graph. They offer a Q/A service and RSS feed subscription for the answers, they provide options to join all sort of groups, to see who looked at your profile recently and more.

What is done today with social graphs?
As you can see the social graph plays a big role in the social network and search arena.  Its usefulness grows significantly as the graph spans multiple social networks and services. Looking at multiple initiatives online, it seems that the next step is to make this graph smarter and more open. Or at the least to make it machine-readable so a computer can find more interesting patterns and information between nodes and links. Sir Tim Berners-Lee, the father of the WWW, now offers

You can also read an interesting response about the semantic graph from Radar Network CEO, Nova Spivac. Radar Network is the company behind twine. Nova speaks about the need for using standard semantic format, such as RDF and OWL, across multiple social network to build the giant global semantic graph. twine itself is a semantic social graph search engine that allows you to build a focused and organized group of resources around a subject matter, called Twine (uppercase T). A Twine is not restricted to web pages only but also allows importing documents, images, videos, and writing notes. The semantic search engine behind it helps to organize the data, automatically suggest tags, and lists related places, people and organizations.  The engines recommend both related twines and twiners (members).

Two more points before moving to alternative social graphs:

* Twitter search  - this is not exactly a social graph search engine but it is a search tool that allows you to search on conversions takking place within Twitter. It allows you to find who said what to whom. It allows you to find and subscribe to feeds for updates by keywords or #hashtags. The TrendingTopic is a snapshot of the current world agenda and sentiment in real-time. Twitter’s following and followers relationships offer an interesting structure that opens up many possibilities for building social graph search engines. You can check two interesting mashup examples using Twitter API like TweetMeme and  TweetWheel.

* There are many attempts for visualizing social graphs and you can see some examples here. As much as it looks fascinating, the usefulness of such a graph for a medium to large scale is in doubt. Yet, it does not reduce the importance of building, maintaining and monitoring such graph in the back-end. Maybe some day in the future someone will come up with a better way to visualize such a graph with multiple options to navigate and focus on certain areas of interest.

Part I recap
We can’t tell yet if the future of searching will be driven by social graphs. Yet we can’t ignore how much links are part of our new daily web routine. We keep on adding links to people, content and other social objects. After looking at some of the existing solutions that leverage capturing and using the graph data structure, I believe that we will see more use for this technology in the future.

The main use of graphs today are to:
* Add more content through nodes (users) from the graph - interactively
* Add more content through crawling using a graph - automated
* Raising content quality and keeping it clean - automated and interactively
* Discovering and encouraging creation of new relationships (explicit and  implicit) - automated

In the next part I will discuss other ways for building and using the social graph.

This entry was posted on Tuesday, September 9th, 2008 at 10:01 am and is filed under Guest Authors. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

Sphere: Related Content