Social Graph Search Engines - Part II
Social graph search engines - Part II - Alternative graph structures
By Keren Dagan
In the first part of this series I explored existing search applications that are using graphs. In this part I will suggest that there are more ways to build a graph. The objective is to show that different graph structures can help to reveal more data and could solve other search problems.
Alternative ways for building social graphs for search engines:
Jyri Engeström from Jaiku provided an interesting perspective about social objects – a good summary can be found in here: 5 principles for Web 2.0 success. In essence he emphasize the interactions around the social objects are a key for success. Yet, most social graphs today only show people as nodes and friend relationships as arcs. What these graphs can’t show us is how people are connected through online activities when they are not connected (yet) as friends, fans or followers. And when people are connected these graphs can’t show us the depth of the relationships.
What Jyri claims is that what that holds people together and drives the relationship is common interests around the social objects. He mentioned LinkedIn as an example where people connect around Jobs and Organizations. Wine lovers’ social networks connect around wine.
So, if we want to find new meaningful connections we need another type of social graph. The objective is exposing additional connections between individuals on the web and their depth: people are still the nodes but the connections (arcs) are made from the pair: unique social object, and action. Example: in the bloggers community the objects are blogger, blog, post, comment and the actions are commenting, linking, rating, digging, posting a blog reaction, and more. The new social graph now makes connections beyond friends, fans and followers. In this graph more arcs are added between nodes when more actions are taken on shared social objects. In this way more meaningful relationships will be exposed vs. superficial ones. The most significant difference using this approach is that it will expose cases where people take action on social objects but aren’t linked as friends, followers, or fans.
Motivation:
* Within a social network: The main objective of a social network service is to make this graph as rich as possible - the more connections are made, the more interactions between nodes are added and the less likely these “nodes” (people) will find the graph dull. LinkedIn is really trying to make an effort in this arena. I think that this is why they came up with groups. Groups allow more ways for people to actually interact - the LinkedIn Q/A section, in my opinion, seems to be moving in the right direction.
* Across multiple social networks: Having such a graph can also help expose new relationships and information. If this kind of graph can cross multiple social networks (i.e. a social graph search engine) it will fill some of the web cracks where information, connections, and opportunities are lost. You can see some examples of information fragmentation on my blog post about Piecing together Web presence. Most times only sophisticated users can overcome the “communication breakdown” that the web creates today. Most people and organizations have no clue how things are connected beyond trivial links. This kind of graph can help both in searching for implicit connections among nodes or in recommending that an implicit connection become explicit.
Most social graphs looks something like this and show only the trivial links between individuals.
I see two more options for building social graphs:
1. To the “trivial” connections also add associative links that may reveal new connections between individuals. Examples: common bookmarks on delicious, a post’s comment, digg-ing your post.
2. Showing the depth of the interactions/relationships between disconnected individuals. As I mentioned above, by showing multiple actions taken on social objects we can expose hidden meaningful relationships.
The social graph search engine can help in the back end, discovering relationships, and prompt us with further actions recommendation. I can also imagine a query like: Am I connected with someone from somewhere? How are we connected? Is this connection has a meaning? This kind of graph can also be used for content discovery as Delver does. It will provide way more connections then the “trivial links”.
I’m aware of the challenges to get all the information for building this graph, privacy issues and dealing with inconsistent identities. There will be a need to access information all over the web after but looking a the following solutions:
* FriendFeed and how they managed to integrate 43 services to one mashup
* Delver and how they manage to discover information from multiple networks
* New services like Gnip and Mashery that are hard at work to consolidate effort around feeds, APIs and protocols may be also a place to normalize data for an entity.
I’m encouraged and think that building such graph is doable.
September 10th, 2008 at 9:49 am
Keren,
These have been two excellent guest spots. Being in PR I’m always looking at maximizing the web, and particularly Web 2.0, to increase my clients’ brand awareness.
These graphs - and your explanations - make excellent points as to where I can look more at the demographic and relation from post to reader. And PR is all about the relationship.
Thanks for the last two days, look forward to more from you.
September 10th, 2008 at 12:08 pm
Danny,
Thank you very much for both comments. I feel that search engines and specifically social search engines can help overcome the fragmented information, discussions, identities on the web today.
The more distributed it is (a good thing) the bigger the problem. Collecting and building graphs using all sort of links can help in my opinion to solve some of these issues and regain lost opportunities. I’m sure that it will be a big time save for PR and marketers too.
Keren
September 14th, 2008 at 4:57 pm
Hi Keren,
first of all thank you for your posts featuring a very new horizon of web development. There’s no market standard yet for this type of web utility and there are also different thoughts about what “social search” is. For example some sources simply associate social search to browsing user’s profiles on social networks such as facebook or myspace, while other sources look at social search as a way to improve algorithm-based web results thanks to the power of the wisdom of crowds. I believe social search is much closer to this second mode rather than the first one. I also believe that there are 2 major inner components within “search”: i) I have an idea, more or less precise, of what I need to find and I need a tool to deliver me the best available results matching that - this is related to a concept of “productivity” of my search results - ii) I’m a curious person and I want to know new things, I want to use the web also as a tool to help me start new ideas - this is related to a concept of “discovery” of my search results -. Now, if that is true, we can understand that there’s a lack between demand and offer of search results. I will explain. Google has done a very good job, of course, working on delivering good results. They have worked on the “productivity” side of the search equation. But almost nothing is done on the other part. Moreover, also the job on the productivity side is not perfect. And honestly, that’s not Google’s algorithm fault. It’s mostly the user’s fault. Because, as I said above, users have a “more or less precise” idea of what they are searching and the combination of words they use when they submit a search is often not accurate. I think algorithm based search is a necessary and useful platform but still not enough. The “last mile” must be human based. And here’s the link to the concept of social search graph: this is a clear concept but still an unclear web format. How to connect people’s searches and related weblinks in a way that increases the productivity of search results, stimulate the discovery factor and is at the same time easy to use is the real challenge of social search enthusiasts. We at Xoost are trying our own way. Honestly, we have not found yet the perfect solution, but we are already applying some of the concepts that I’m reading in this post. And we are trying day by day to improve our product and to match user’s requirements. The way ahead is long and tricky but very stimulating. Cheers, Felix