[IBM Research] SOSA: Breaking the Chains of Social Silos through Semantic Middleware
Social Ontology and Semantic Actions
This paper introduces Social Ontology and Semantic Actions (SOSA), a decentralized middleware framework designed to achieve interoperability across siloed social networking platforms. By utilizing a "Contact Graph Object" and relationship ontologies, it enables cross-platform contact consolidation and the automated execution of enterprise services (Semantic Actions) based on derived social relationships.
TL;DR
The Social Ontology and Semantic Actions (SOSA) framework is a decentralized architecture that allows users to own their social graphs and port them across platforms. By introducing Contact Graph Objects and Semantic Actions, SOSA enables users to consolidate identities from Facebook, LinkedIn, and Gmail, and trigger enterprise tasks (like sending a "Request for Quote") based on inferred social relationships.
Background: The "Information Silo" Problem
In the current digital landscape, social networks act as walled gardens. Users invest massive energy into building contact lists on one platform, only to find that data trapped when they move to another. As noted by the authors, "People are getting sick of registering and re-declaring their friends on every site."
From an enterprise perspective, the problem is worse: businesses cannot define their own specialized relationships (like "Supplier" or "Verified Client") within the rigid frameworks of consumer platforms like Facebook. SOSA was conceived to solve this by moving the social logic into a portable middleware layer.
Methodology: How SOSA Works
The core of SOSA lies in its ability to treat "relationships" as actionable data points rather than just static links.
1. The Contact Graph Object (CGO)
Instead of a single profile, a CGO acts as an aggregator. It uses Conflict Links (to identify duplicate identities via Levenshtein distance algorithms) and Fact Links (to store attributes like birthdays or roles).
2. Relationship Axioms & Ontologies
SOSA doesn't just store "Friend" status. It uses an extensible ontology. If you define a "Brother" relationship, the system can use inference to include that person in a "Family" query automatically. This removes the need for manual group management.
Figure 1: The SOSA architectural framework, showing the interaction between the Social Query Runtime and the Graph Store.
3. Semantic Actions: The "How" of Integration
This is the most innovative part: Semantic Actions. A user can bind a RESTful service to a relationship. For example: email(<siblings>, <wish_list_url>). When the "siblings" axiom is triggered, the system dynamically resolves the contacts and executes the web task.
Figure 2: The User Interface for defining social axioms within the SOSA middleware.
Real-World Scenario: The Portable Social Graph
The paper illustrates a powerful use case:
- Phase 1: Joe consolidates his contacts from Facebook and LinkedIn on his phone using SOSA.
- Phase 2: Joe gets engaged. He exports his "Family" graph to an e-commerce site (Shop.com) to share a wish list.
- Phase 3: At work, Joe exports his "B2BPartner" graph to a government server to issue a tender.
In all cases, the social data remains under Joe's control and is simply "imported" into the local SOSA component of the service he is currently using.
Experimental Insight & Results
While primarily a framework proposal, SOSA addresses the critical metrics of Transferability and Inference. By using SPARQL queries against a Social Graph Store, the authors demonstrate that memberships can be computed on-the-fly, reducing the manual overhead of contact management by nearly 100% for derivative groups.
Figure 3: An example of a Semantic Action Template binding a social group to a distributed web task.
Critical Analysis & Conclusion
Takeaway
SOSA represents a significant step toward the "Social Web" envisioning a world where social data is as portable as an email address. Its strength lies in its Enterprise readiness—allowing companies to host their own social components behind firewalls while remaining compatible with public APIs.
Limitations
The paper acknowledges that Relationship Ontology Life Cycle Management (how relationships change over time) and Performance Tuning for large-scale graphs remain areas for future research. Additionally, the reliance on users to manually reconcile "Conflict Links" might be a friction point in very large networks.
Future Outlook
As we move toward a more decentralized internet (Web3), the principles in SOSA—specifically the de-coupling of the social graph from the application layer—will likely become the standard for how we interact with digital services.
