EasyGo: Orchestrating Social Circles through Semantic Web Integration

TL;DR

"EasyGo" is a Semantic Web framework that harvests event data from fragmented online marketplaces (like StubHub and Ticketmaster) to build a unified social ecosystem. By integrating heterogeneous data into an ontology and using LDA-based topic modeling, it enables users to find theater, sports, and concert events, form purchase groups, and transition online interest into offline friendships.

Context & Motivation: The Fragmented Event Universe

Most event-based social networks are "walled gardens"—systems built from the top-down where you only see what the platform owner permits. However, the real world of events is messy and scattered across dozens of ticket vendors.

The researchers identified a critical gap: How can we allow users to self-organize social circles around any event on the web? The core challenge isn't just finding the events, but resolving the "Heterogeneity Conflict"—where one site lists a "Venue" and another a "Location," or one refers to "Taylor Swift" while another says "Taylor Swift featuring Ed Sheeran."

Methodology: The Semantic Glue

The authors propose a system architecture (Fig. 1) centered around a Triple Store, acting as the knowledge base for the entire social network.

1. Data Transformation (Karma & Scrapy)

Using Scrapy for crawling and the Karma semantic tool, the system turns raw JSON/HTML data into structured RDF triples. This allows the system to treat a StubHub listing and a Ticketmaster listing as comparable semantic entities (Fig. 2).

2. The Hybrid Matching Algorithm

To solve the naming mismatch problem, the authors use a dual-weighted approach:

• Syntactic Similarity: Using Levenshtein distance to catch spelling overlaps.
• Semantic Similarity: Using WordNet synsets and Jaccard similarity to realize that "Venue" and "Location" mean the same thing.

For instance-level matching (identifying if two listings are the same concert), they apply the Smith-Waterman Similarity, which is robust for local sequence alignments—essential for catching artist names within long event titles.

Recommending Connection: The Role of LDA

Instead of just recommending friends based on mutual friends (the Facebook model), EasyGo uses Latent Dirichlet Allocation (LDA).

• Input: User profiles (extracted from Facebook) and event descriptions.
• Process: Topics are extracted as latent variables.
• Output: A recommendation score based on the Cosine Similarity between the user’s topic vector and the event’s topic vector.

Experimental Results & Application

The framework was realized in the EasyGo web application. It successfully demonstrated that users could:

1. Find the cheapest deals across multiple platforms.
2. Join or create "Groups" to share delivery and service fees.
3. Automate the formation of social ties before the physical event takes place.

Critical Insight & Future Work

The true value of this work lies in its Semantic Interoperability. By moving away from proprietary databases to a Triple Store, the "Social" part of the network becomes an emergent property of the data rather than a hard-coded feature.

Limitations: Currently, the ontology refresh rate is manual (once per day). In the high-stakes world of ticket sales, where prices fluctuate by the minute, this is a bottleneck. The authors' future plan to use machine learning to predict optimal update frequencies is a necessary evolution for this system to survive in a commercial environment.

Conclusion

EasyGo proves that the Semantic Web is not just a theoretical exercise for researchers—it is a powerful tool for data mashups that can drive real-world social behavior and economic savings.