Scoring Model based on Friend-network Relationships for a Social Aware Search Application
As for 2016, the E-commerce sector continues to grow at a fast pace but still represents only about 10% of all retail sales (depending on the market sector). This implies a great potential to grow and we can expect that its relevance will still increase in the near future. This fact is known to companies which must optimize their in-store experiences with connections between online and offline worlds.
One such possible optimization concerns with the improvement of the customer search experience in an online shop. This enhancement is based on the personalization of the search requests to which the system adds the customer's social context. The primary target is to optimize customer lifetime values (LTV) which improve the customer experience on one hand and increases the company’s revenue on the other.
The "Social-aware search" concept deals with this kind of optimization. Its objective is to gather data from the customer's social-network account (like personal information and social connections) in order to construct a user context which is then used to create a personalized search experience. This includes the contextualization of the input search keywords, a personalized search results ranking or tailored product recommendations.
The objective of this task is to conduct some research on the different scoring models and algorithms used to rank social media in centralized social networks. It shall be analyzed, how the existing scoring models and algorithms can be applied together in order to develop a new scoring model suitable for improving the customer’s search experience. The resulting model should leverage social network metadata like social connection types in order to create groups with different ranking weights, or the system's metadata itself, allowing to take into account properties like data-aging or click-through rates.
All this information should be used to create a weighted social graph and a data-base of linked data and applied in a demonstrator providing results ranking functionality and a set of tailored recommendations.
The master thesis includes a State of the Art analysis beforehand and the implementation of a demonstrator with a scoring model. The primary target of this thesis is to show an objective improvement to the processing of the search results provided to users. This should be demonstrated through several test cases.