# Collection Ranking and Selection for Federated Entity Search ## queries.txt Queries from the [2010] (http://km.aifb.kit.edu/ws/semsearch10/) and [2011] (http://semsearch.yahoo.com) editions of the Semantic Search Challenge. Line numbers correspond to query IDs. (2010 queries: #1 ... #92; 2011 queries: #93 ... #142.) Queries were processed using the [Yahoo! Spelling Suggestion API] (http://developer.yahoo.com/search/web/V1/spellingSuggestion.html). This affected query IDs: #2, #10, #43, #66, #67, #108, #112, #114, #120. The file contains all queries, even the ones that do not have any relevant results in the collection ranking qrels. ## topdomains.txt The list of top 100 domains used as distributed collections; this corresponds to the "BTC" setting in the paper. The "BTC\DBpedia" setting uses the same list of domains except dbpedia.org. ## dbpedia-splits The `dbpedia-uri-X.txt` files, X=0..99, contain the URIs contained in each of the splits. X is used as the identifer for the given subset in the collection-ranking qrels. ## qrels.collection-ranking Collection ranking qrels. A collection is considered relevant if it contains at least one relevant entity (i.e., with relevance level >0). For graded relevance, the gain for each collection is set to the number of relevant documents the collection contains. - `topdomains.qrels`: top 100 domains, i.e., the "BTC" setting in the paper. - `topdomains-dbpedia.qrels`: top 100 domains minus DBpedia, i.e., the "BTC/DBpedia" setting in the paper. - `dbpedia.qrels`: DBpedia only. ## Reference If using these resources, please cite our paper:
@inproceedings{Balog:2012:CRS, author = {Krisztian Balog and Robert Neumayer and Kjetil N{\o}rv{\aa}g}, title = {Collection Ranking and Selection for Federated Entity Search}, booktitle = {Proceedings of the 19th International Symposium on String Processing and Information Retrieval (SPIRE 2012)}, publisher = {Springer}, pages = {73-85}, year = {2012} }## Contact In case of questions, feel free to contact us: - Krisztian Balog