» What is ios?
Instant Overview Search, for short ios, is a "search-as-you-type" functionality for web searching which provides an overview of the results as the user types his query letter-by-letter. Specifically, for the most frequent queries it shows instantly the top hits and various other kinds of precomputed aggregated information that provide the user with a better overview of the search space. This method is independent of the kind of precomputed information, hence it can be adopted for a plethora of applications (e.g. results clustering, metadata-based groupings, entity mining, query recommendations, semantic-based enhancements, etc).
At present, we investigate how to exploit the Semantic Web technologies for enriching the entities that have been identified in the web search results. Specifically, we study how the Linked Open Data cloud can be exploited for providing further information about the entities that lie in the search results of both general purpose web search engines (running prototype 5) and vertical search scenarios (running prototype 6).
Instant Overview Search, for short ios, is a "search-as-you-type" functionality for web searching which provides an overview of the results as the user types his query letter-by-letter. Specifically, for the most frequent queries it shows instantly the top hits and various other kinds of precomputed aggregated information that provide the user with a better overview of the search space. This method is independent of the kind of precomputed information, hence it can be adopted for a plethora of applications (e.g. results clustering, metadata-based groupings, entity mining, query recommendations, semantic-based enhancements, etc).
At present, we investigate how to exploit the Semantic Web technologies for enriching the entities that have been identified in the web search results. Specifically, we study how the Linked Open Data cloud can be exploited for providing further information about the entities that lie in the search results of both general purpose web search engines (running prototype 5) and vertical search scenarios (running prototype 6).
» Running prototypes
1) A meta-search engine offering instant clustering of the snippets of the top hits. The current deployment contains precomputed information for 20,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the "fresh" results and their cluster label tree. The prototype also supports faceted search-like restriction of the answer, i.e. the user is able to gradually select clusters and refine the answer set accordingly. For the clustering of the top results, we adopt the algorithm NM-STC.
2) A standalone web search engine offering instant meta-data based grouping of the top hits. In particular, user can limit the search space according to various dynamic taxonomies (e.g. filetype, date, domain, etc.). The current demo contains precomputed information for about 100 queries. In case user does not get any suggestion for the current input, he is able to submit the query. In that case, he is redirected at mitos web search engine. You can find more information about dynamic taxonomies here.
3) A meta-search engine offering instant entity mining of the contents of the top hits. The current deployment contains precomputed information for about 2,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the results and a categorized list with the discovered entities (of either the snippets or the full contents of the results). You can find more information about web searching with entity mining at query time here.
4) A meta-search engine offering both instant clustering of the snippets and instant entity mining of the contents of the top hits. The current deployment contains precomputed information for about 2,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the cluster label tree and a categorized list with the discovered entities (of either the snippets or the full contents of the results).
5) A meta-search engine offering "semantically enriched" entity mining of the top hits. Specifically, user is able to retrieve more information about a discovered entity by simply clicking the icon
next to entity's label.
In that case and according to the category of the selected entity,
a SPARQL query is sent trying to fetch more information about that entity from the Linked Open Data cloud,
exploiting various LOD datasets that offer access through SPARQL endpoints (like DBpedia and FactForge).
6) Case Study: "Fisheries and Aquaculture publications". Apart from the case of general purpose web searching, we have started investigating the "semantically enriched" entity mining approach in vertical search scenarios. One of this is the domain of FAO (Food and Agriculture Organization) publications about fisheries and aquaculture. The underlying keyword search system is the FIGIS search component. As concern entity mining, we identify the following additional relevant categories: Countries, Water Areas, Regional Fisheries Bodies, and Marine Species. Furthermore, user is able to retrieve more information about a discovered entity from the Fisheries Linked Open Data (FLOD) RDF dataset (FLOD is exposed via a public SPARQL endpoint).
1) A meta-search engine offering instant clustering of the snippets of the top hits. The current deployment contains precomputed information for 20,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the "fresh" results and their cluster label tree. The prototype also supports faceted search-like restriction of the answer, i.e. the user is able to gradually select clusters and refine the answer set accordingly. For the clustering of the top results, we adopt the algorithm NM-STC.
2) A standalone web search engine offering instant meta-data based grouping of the top hits. In particular, user can limit the search space according to various dynamic taxonomies (e.g. filetype, date, domain, etc.). The current demo contains precomputed information for about 100 queries. In case user does not get any suggestion for the current input, he is able to submit the query. In that case, he is redirected at mitos web search engine. You can find more information about dynamic taxonomies here.
3) A meta-search engine offering instant entity mining of the contents of the top hits. The current deployment contains precomputed information for about 2,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the results and a categorized list with the discovered entities (of either the snippets or the full contents of the results). You can find more information about web searching with entity mining at query time here.
4) A meta-search engine offering both instant clustering of the snippets and instant entity mining of the contents of the top hits. The current deployment contains precomputed information for about 2,000 queries. In case user does not get any suggestion for the current input, he is able to submit the query and get the cluster label tree and a categorized list with the discovered entities (of either the snippets or the full contents of the results).
5) A meta-search engine offering "semantically enriched" entity mining of the top hits. Specifically, user is able to retrieve more information about a discovered entity by simply clicking the icon
next to entity's label.
In that case and according to the category of the selected entity,
a SPARQL query is sent trying to fetch more information about that entity from the Linked Open Data cloud,
exploiting various LOD datasets that offer access through SPARQL endpoints (like DBpedia and FactForge).
6) Case Study: "Fisheries and Aquaculture publications". Apart from the case of general purpose web searching, we have started investigating the "semantically enriched" entity mining approach in vertical search scenarios. One of this is the domain of FAO (Food and Agriculture Organization) publications about fisheries and aquaculture. The underlying keyword search system is the FIGIS search component. As concern entity mining, we identify the following additional relevant categories: Countries, Water Areas, Regional Fisheries Bodies, and Marine Species. Furthermore, user is able to retrieve more information about a discovered entity from the Fisheries Linked Open Data (FLOD) RDF dataset (FLOD is exposed via a public SPARQL endpoint).
» Publications
1) P. Fafalios and Y. Tzitzikas
"Exploiting Available Memory and Disk for Scalable Instant Overview Search"
12th International Conference on Web Information System Engineering, WISE 2011, Sydney, Australia, October 2011
(pdf • bib)
2) P. Fafalios, I. Kitsos and Y. Tzitzikas
"Scalable, Flexible and Generic Instant Overview Search"
Demo Paper, Proceedings of the 21st International Conference on World Wide Web, WWW 2012, Lyon, France, April 2012.
(pdf • ppt • bib)
3) P. Fafalios, I. Kitsos, Y. Marketakis, C. Baldassarre, M. Salampasis and Y. Tzitzikas
"Web Searching with Entity Mining at Query Time"
Proceedings of the 5th Information Retrieval Facility Conference, IRF 2012, Vienna, July 2012.
(pdf • bib)
1) P. Fafalios and Y. Tzitzikas
"Exploiting Available Memory and Disk for Scalable Instant Overview Search"
12th International Conference on Web Information System Engineering, WISE 2011, Sydney, Australia, October 2011
(pdf • bib)
2) P. Fafalios, I. Kitsos and Y. Tzitzikas
"Scalable, Flexible and Generic Instant Overview Search"
Demo Paper, Proceedings of the 21st International Conference on World Wide Web, WWW 2012, Lyon, France, April 2012.
(pdf • ppt • bib)
3) P. Fafalios, I. Kitsos, Y. Marketakis, C. Baldassarre, M. Salampasis and Y. Tzitzikas
"Web Searching with Entity Mining at Query Time"
Proceedings of the 5th Information Retrieval Facility Conference, IRF 2012, Vienna, July 2012.
(pdf • bib)
» Contact
This system was developed in the scope of Pavlos Fafalios master's thesis under the supervision of Yannis Tzitzikas, assistant professor at the University of Crete, and was partially supported by the EU project iMarine (FP7-283644).
This system was developed in the scope of Pavlos Fafalios master's thesis under the supervision of Yannis Tzitzikas, assistant professor at the University of Crete, and was partially supported by the EU project iMarine (FP7-283644).
Pavlos Fafalios
Computer Science Department, University of Crete, Greece
Information Systems Laboratory, Institute of Computer Science, Foundation for Research and Technology - Hellas (FORTH), Greece
Email: fafalios@csd.uoc.gr
Web page: http://www.ics.forth.gr/~fafalios/
Computer Science Department, University of Crete, Greece
Information Systems Laboratory, Institute of Computer Science, Foundation for Research and Technology - Hellas (FORTH), Greece
Email: fafalios@csd.uoc.gr
Web page: http://www.ics.forth.gr/~fafalios/