Sarthak Ahuja (2012088) Saumya jain (2012089) NEWSS – Social Search Sarthak Ahuja (2012088) Saumya jain (2012089)

Social Search and folksonomy A behavior of retrieving and searching on a social searching engine that mainly searches user-generated content such as news, videos and images related search queries on social media Social Information Retrieval Folksonomy: Social resource sharing systems all use the same kind of lightweight knowledge representation Taxonomy + folks conceptual structures created by the people. There is currently a number of research work performed in the area of bridging the gap between Information Retrieval (IR) and Online Social Networks (OSN). This is mainly done by enhancing the IR process with information coming from social networks, a process called Social Information Retrieval (SIR)

Formal Model for Folksonomies Tuple: (U, T, R, Y) U: Users, described by IDs T: Tags, arbitrary strings R: Resources, depend on the type of system implemented Y: ternary relation Y ⊆ U x T x R

Adapted PageRank STEP 1: V: disjoint union of tags, users, resources E: Co-occurrences of these three sets (undirected, weighted edges) Hypergraph between sets of users, tags, resources F = (U, T, R, Y) undirected, weighted, tripaprtite graph G = (V, E) Tripartite: The tripartite structure of the graph can be exploited later for an efficient storage of the – sparse – adjacency matrix and the implementation of the weight-spreading iteration in the FolkRank algorithm.)

Adapted PageRank STEP 2: Apply PageRank on this graph. Takes edge weights into account Basic notion: a resource which is tagged with important tags by important users becomes important itself (symmetric for tags and users) Random surfer model w ← aw + ßAw + γp A: row-stochastic version of G p: preference vector a, ß, γ ∈ [0, 1] with a + ß + γ =1 a regulates speed of convergence proportion between ß and γ controls the influence of the preference vector Like PageRank, we employ the random surfer model, a notion of importance for web pages that is based on the idea that an idealized random web surfer normally follows hyperlinks, but from time to time random a regulates speed of convergence, while the proportion between ß and γ controls the influence of the preference vector ly jumps to a new webpage without following a link.

FolkRank (Topic specific ranking) Compare resulting ranking with and without preference vector Preference vector used to determine the topic. Give high weight to tags/users/resources Weight vector given by: w1 – w0 w ← aw + ßAw + γp w0 : ß =1 w1: ß <1

Comparing FolkRank with Adapted PageRank

References Information retrieval in folksonomies: Search and ranking, A Hotho, R Jäschke, G Stumme - 2006 – Springer Social networks and information retrieval, how are they converging? Mohamed Reda Bouadjenek – 2016 – Elsevier

THANK YOU!