1 GROUPER A DYNAMIC CLUSTERING INTERFACE TO WEB

1 GROUPER: A DYNAMIC CLUSTERING INTERFACE TO WEB SEARCH RESULTS Erdem Sarıgil - 21000089 Oğuz Yılmaz - 21000082

Grouper 2 Interface to the results of the Husky. Search Dynamically groups the search results into clusters using Suffix Tree Clustering Algorithm (STC) The goal make search engine results easy to browse by clustering them Grouper receives hit from different engines, and only looks at the top hits from each search engine

Post-retrieval Clustering 3 Based on the returned document set Superior results than pre-retrieval clustering Some key requirements: � Coherent Clusters � Efficiently Browsable � Speed Algorithmic Speed Snippet-Tolerance

Suffix Tree Clustering (STC) 4 Linear time clustering algorithm STC has three logical steps: � Document cleaning � Identifying base clusters using a suffix tree � Merging these base clusters into clusters STC has several novel characteristics: � Overlapping clusters � Bag-of-words � Well suited for Web document clustering � Robust in such “noisy” situations

User Interface 5

User Interface (cont’d) 6

7 Making the Clusters Easy to Browse Three heuristic to identify redundant phases: 1. 2. 3. Word Overlap Sub- and Super- Strings Most General Phase with Low Coverage

Speeeeed 8 Quality Search Time the vice president of Quality OR Time Quality vice president

Coherent Clusters 9

Comparison 10 Number of documents followed Time Spent Click Distance

Comparison (cont’d) 11

12