Its all about the Buddy System
Recently I have been engaging a few of my industry cohorts on the topic of social search engines and the future of search in general. In the coming weeks I will be publishing some of these discussions which for me lead to some hybrid of algorithmic, performance metric based and human enhanced ranking signals.
One of the problems facing those that believe in a pure social powered search engine is the fact that over time, once the novelty wears off, many users will be less inclined to actively be involved and so the few (in the form of power users) would be creating indexes/SERPs for the many. One way of dealing with this (and spam) would be to have a form of personalized (trusted) network search approach
But how do you make creating networks and accessing ranking signal easier? Yahoo seems to have a plan.
In a patent I came across yesterday, I would seem the folks at Yahoo! are looking to address such shortcomings in a personalized/social search sphere.
Systems and methods for establishing or maintaining a personalized trusted social network
As part of the example given they highlight their My Web 2.0 web services in what they call;
an example of a web-scale social search engine that enables users to find, use, share and expand human knowledge. It allows users to save and tag Web objects, allowing for browsing and searching of objects, as well as sharing Web objects within a personalized community or to the public. Further, the MY WEB 2.0 Web service provides scoped searches within a user's trusted social network (e.g., friends and friends of friends). As a consequence, the search results are personalized and spam-filtered by trusted networks.|
The trouble being the number of hoops one has to jump through when personalizing the experience and including others in your trusted network;
|However, to complete the (above) process, the user must expend a considerable amount of time to collect the requisite data about the members of the user's social network. Considerable effort must also be expended to populate the databases with saved and tagged objects. As such, a user, including the members of the user's social network, can be initially discouraged from seeking the advantages of personalized web services.|
The Quest for a Social Search Engine
This particular patent is a continuation of some other patents that also seem to be part of the My Web offering and referenced in this document;
Search systems and methods with integration of user annotations
Search systems and methods with integration of user annotations
Systems and methods for collecting user annotations
In essence this is a methodology for building a trusted social network with the least amount of input required from the user. In terms that were familiar with, users can define social networks and subsets of networks that can be accessed via invite, association etc
This seemingly can include users of other networks, professional associations and the like. Furthermore looking at tagged items of a given user further leverages potential trusted members for a personalized network;
|For example, if a given user is a subscriber to an online service (such as, a blog), an RSS feed for the online service can be parsed to identify the owner (e.g., trusted source) of the online service as well as any other contributors or the like, including individuals or entities that have contributed to the online service by adding content items (e.g., comments) and/or attaching microcontent items (e.g., related articles). The identified individuals and/or entities (e.g., the owner and other contributors) are added as members to the user profile as trusted sources for the given user.|
Once again one can also look at other social services in the users profile to further mine trusted members of the users social network. The main concept being that there is a perceived value for people that are in your other networks or authors of tagged content. One could also look at the niche types of users on existing social networks to find relevant signals.
Also proposed is not only seeking previous trusted sources via your social networks but also looking at the tagged micro-content of trusted members from other websites to create further meta-data potential personalization.
Hey Buddy, whats your association?
They describe two types of potential users in a trusted network;
- a friend,
- family member,
- non-profit organizations,
- government agencies,
- activist groups,
- educational institutions,
- news sources,
- file-sharing communities,
- professional associations
These are not mutually exclusive though and can all be looked for creating a personalized social network. The use must invite or identify buddies and specify associations with entities.
Now we can look at any annotations that have been made on content items within the trusted network. Annotations can include any descriptive or evaluative meta-data relating to content items such as thumbs up/down, keywords associated with the content, free text descriptions and so forth.
One example usage includes annotating search result content items as well as actual web pages visited. This data can then be used to personalize subsequent search results. When presenting the results items that have previous annotations can be highlighted with a link to the annotation in the results with the listing. The same goes for any evaluative tags such as a positive/negative rating (think StumbleUpon thumbs). Such judgement data can even be used to refine rankings in the personalized search results in a positive or negative manner depending on the ratings.
The user can alternatively decide to search annotations themselves in addition to or instead of the page content. The data for search personalization can include; navigation history, tags, annotations and more.
Personalized Network Search
For implementation, I cant help but think that deeper adoption of the OpenID platform could be extremely useful for not only this implementation, but the future of social search in general. I really havent played around with Yahoos - My Web 2.0 - much to be honest, but have loaded up the Y toolbar and will see what it has to offer; more on that another day. Ultimately this seems to have some interesting potential for unlocking the code to producing a social search experience.
In the end, to avoid spam, there certainly needs to be a way to create trusted networks whom alone can affect the ranking of documents in the search results. This is a form of expanded personalized search that can work well if blended with traditional personalized search signals (user performance metrics).
If you have a vision for the perfect Social Search Engine, let me know as I am currently working on a few posts in this area
Until next time; stay tuned.