A Survey of Exploratory Search Systems Based on LOD Resources

International Conference on Computing and Informatics (ICOCI) Istanbul -2015
Karwan Jacksi
PhD Student
Computer Science Dept.
University of Zakho – Kurdistan Region of Iraq
And
School of Computing and Technology
Eastern Mediterranean University – Cyprus
A Survey of Exploratory Search Systems Based
on LOD Resources

Agenda
• WWW
• Search Strategies
• Semantic Web
• Linked Data
• Linked data browsers
• Linked data recommenders
• Linked data based exploratory search systems
• Discussions
• Computing Semantic Similarity
• Database technique
2015.08.12Karwan Jacksi 2

World Wide Web (WWW)
• WWW made the process of uploading data easy for all users
– The amount of data grew in an unprecedented manner, which has given the
internet users opportunity to access information and services effortlessly.
– This data has transformed the Web to a huge semi-structured database.
• Thus, the new challenge to overcome is to efficiently retrieve relevant
information
• Even though search engines are generally employed to locate information
on the web, to obtain high quality results, efficient searching skills are
required.

Search Strategies
• (Marchionini, 2006) has distinguished between two categories of search
strategies:
1. Lookup search
– known as keyword-based search,
– known-item search
– most frequently used one in the syntactic Web
2. Exploratory search
– a special type of information seeking
– the domains of the goals are unknown and the goal itself is not known.
– user interest focuses on learning and investigation rather than query answering and
information retrievals
– learning:
• understanding new ideas and concepts, knowledge achievement and information
assessment,
– investigation:
• analysis, composition and evaluation
Lookup Exploratory
Learn
Investigate

Syntactic vs Semantic Web
• The Syntactic retrieval of information is restricted to keywords
– to answer user queries, search engines search for keywords which do not
retrieve high quality results
• To improve this, annotations can be added to the existing Web contents
• Hence, the Semantic Web is formulated

Semantic Web
• the next generation of the WWW
• provides data well defined meanings so that it’s understandable and processable
by machines
• combines Resource Description Framework (RDF), Web Ontology Language
(OWL), and Extensible Markup Language (XML) technologies to facilitate
replacing the content of the current Web
– RDF is W3C standard
– RDF is known as the HTML of the Semantic Web and it is purely an XML language
– RDF is a standard model for interchanging of data on the Web and it simply describes
resources on the Web in the form of <subject, predicate, object> expressions called
triples
– Ontology is a vital components of the Semantic Web infrastructure
• OWL and RDF Schema (RDFS) are the data representation models recommended by W3C to
provide elements for the description of ontologies
• these technologies give machines a way to read and understand the meaning of
data. Consequently, automated information processing capability is given to
search systems.

Linked Data
• Linked Data is “the Semantic Web done right” (Berners-Lee, 2006)
• Berners-Lee remarks that the Semantic Web is not only about annotating
data on the Web, but it is also about linking data with each other so that
people and machines can discover the Web of data.
• Linked data provides a way to find other related data when only a subset is
given.
• The Linked Data is a term that refers to a set of steps for publishing and
linking structured data on the Web.

Linked data browsers
• One of the first generation of tools designed to explore linked data was
semantic browsers.
• Before their existence it was necessary to read the serialized RDF files to
discover and understand the data.
• There are:
– Text-based browsers
– Visualization based browsers
– Faceted browsers
– Other browsing paradigms

Linked data recommenders
• Several linked data exploration and discovery approaches are based on
similarity or relatedness measures
• Their objective is to identify the resources that are the most similar or
related to an initial resource or resources’ set
• Measuring similarities thanks to linked data datasets is a relatively new
concern
• Examples:
– MORE (2010) is a DBpedia-based movie recommender accessible through a
Facebook application
– LastFM (2010) music domain for computing similarities between all the Bands
and Musical Artists instances contained in DBpedia.

Linked Data (LD) based Exploratory Search Systems (ESS)
• Yovisto (2009)
– A search engine focused in academic lectures and conferences’ videos.
– Provides explorative search for video data using Semantic Web technology and
linked data
– Searching within the content of videos is possible with Yovisto through its time
based video index
– The approach utilizes DBpedia as a resource
– DBpedia provides a large amount of
information about each of its entities,
• retrieving all relevant information is
computationally intensive
– Hence, a set of heuristics, based on statistical
and structural features of the DBpedia’s RDF
graph, are developed to determine the most
relevant information to the users query.

LD based ESS Cont’d
• Semantic Wonder Cloud (SWOC) (2010)
– A linked data based exploratory search system for DBpedia.
– A hybrid ranking algorithm, called DBpedia Ranker, is used to rank resources in
the DBpedia. This algorithm is used to calculate the similarity of DBpedia
resources with respect to an initial node.
– External sources such as Web search engines and social tagging systems is also
used to calculate a similarity value for each of the resource pairs touched
throughout the discovery.

• Lookup Explore Discover (LED) (2010)
– An exploratory search system that aims at improving the search process by
allowing users to accurately explore information related to a query
– It depends on DBpedia dataset to discover the semantics of the query keywords
– DBpedia lookup service is used to return a set of DBpedia resources
– Result is a set of semantically related resources of the form of a tags cloud
– It uses several external sources such as Google, Bing, Yahoo and micro-blogging
services such as Twitter
– DBpediaRanker is used to
rank resources

• Aemoo (2012)
– a Web application that uses Encyclopedic Knowledge Pattern (EKP) to support its
exploratory search system.
– the system exploits DBpedia first to resolve users’ query, then external sources
such Google News and Twitter
– For a better relevant and reasonable information, resources are filtered using
the EKP
– Aemoo offers Curiosity function to display the knowledge that is not presented
due to the EKP filtration

• Seevl (2012)
– a Linked Data based application for musical discovery.
– The Web is used by Seevl to mine music connections so that context, search and
discovery are brought for the music a user likes.
– Multiple Web resources are collected and unified as Linked Data.
• As a result, a huge graph of musical entities, such as Artists, Labels, Bands, etc., is built.
– On top of this collected data, further services, for instance search and
recommendations, are provided.

• Linked Jazz (2013)
– an exploratory system that reveals the relationships of the jazz community.
– The system aims at improving the discovery of digital cultural heritage materials
and enriching the semantics describing them examining the power of LOD
technology.
– Dataset is created using bibliographic name authority and DBpedia
– Network Visualization tool is used to explore the created dataset which provides
a number of modes, such as Fixed, Free, Similar and Dynamic modes

• Discovery Hub (2014)
– The platform uses DBpedia datasets for resource selection
– It addresses DBpedia SPARQL endpoints remotely without any pre-processing
requirements.
– The system depends on an extension of semantic spreading activation to typed
graphs of the Web of Data formality and combined it with a graph sampling
method to calculate result without the need of in advance processing.

• inWalk (2014)
– an interactive Web application for exploration of linked data based on the
concept of inCloud
– An inCloud is a high level graph of thematically connected vertices, where
vertices of the graph represent clusters of related linked data and edges stand
for relations of proximity between vertices.
– inCloud is constructed using hierarchical clustering algorithm HCf+

System Name Yovisto SWOC LED Aemoo
Release Date 2009 2010 2010 2012
Web Address www.yovisto.com
sisinflab.poliba.it/semantic-
wonder-cloud
sisinflab.poliba.it/led wit.istc.cnr.it/aemoo
Main Data DBpedia EN+DE Dbpedia EN Dbpedia EN DBpedia EN
Auxiliary Data No
Search Engines & Tagging
Systems
Search Engines & Tagging
Systems
External services
Query Model Keyword search Dbpedia lookup Keyword search Lookup
Matching String-match Direct match (lookup) String matching Direct match (lookup)
Domain in (Purpose) Academic Videos IT Domain ICT General
Database Method Freebase Parallax SWOC Storage LED Storage
Knowledge Pattern (KP)
Repository Manager
Databse Purpose map queries to entities
Stores popularity & similarity
values for pairs of resources
Stores results computed by
Ranker
responsible for storage,
indexing & fetching of KPs
Principal Layout Query suggestions Graph Tags cloud Graph
Results Explanations No No No Wikipedia-based
Breadcrumb Sessions & registeration Sessions Sessions Sessions
Algorithms Set of heuristics DBpedia Ranker DBpedia Ranker EKP filtered view
Ranking Yes Yes, graph size Yes No
Offline Processing Yes Yes, similarity of pairs Yes Yes , EKP part
API RDF triple-store No Yes, RESTful Yes, RESTful
Faceted Navigation Yes No No No
User Interface HTML Flash based HTML HTML

System Name Seevl Discovery Hub Linked Jazz inWalk
Release Date 2012 2013 2013 2014
Web Address play.seevl.fm http://discoveryhub.co https://linkedjazz.org http://islab.di.unimi.it/inwalk
Main Data
Dbpedia, Freebase &
MusicBrainz
DBpedia EN, FR, IT Dbpedia -> Linked Jazz DB Freebase & Twitter
Auxiliary Data No No No No
Query Model Lookup Dbpedia Lookup Manual selection from a list Lookup & selection
Matching Direct match (lookup) Direct match (lookup) Selection Direct match (lookup)
Domain in (Purpose) Music General Jazz Musicians Athletes or Twitter News
Database Method
Scalable RDF/ NoSQL storage
by OpenLink Virtuoso
Virtuoso, MySQL Linked Jazz Name Directory inWalk repository
Databse Purpose Scalability User account & follows
Stores individuals
represented by literal triples
Provides a high-level view of
relevant linked data
Principal Layout List List Graph Graph
Results Explanations Shared properties Yes, Text & Graph Yes, Interview Transcripts No
Breadcrumb Sessions & registeration Sessions & registeration Sessions Sessions
Algorithms LDSD, DBrec algorithm
Semantic spreading
activation
Mapping & Curator Tool & the
Transcript Analyzer
HCf⁺ clustering algorithm
Ranking Yes Yes No No
Offline Processing Yes No, on the fly Yes, Linked Jazz DB Yes
API
Yes, Content negotiation JSON-
LD
No Yes, JSON, RDF & GEXF files No
Faceted Navigation Yes Yes No No
User Interface HTML, ajax HTML HTML5 + jQuery HTML5 + JS

Discussion
1. All systems except for inWalk, use DBpedia as main datasets to extract
data from
2. SWOC, LED and Aemoo additionally utilize external services such as
search engines and tagging systems to compute associations between
DBpedia nodes.
3. DBpedia lookup service has been notably used by most of the systems as
query model.
4. Aemoo and Discovery Hub are multi-purpose search systems where all
types of information can be retrieved without directing to any specific
domain.
5. All systems offer breadcrumb feature (history) using sessions which helps
users easily retract to their previous state allowing comparison and
analysis of the results.
6. Yovisto, Seevl and Discovery Hub offer registration feature,
1. this keeps search history for registered users at the server side.

Discussion Cont’d
7. Aemoo, Seevl, Discovery Hub and Linked Jazz offer explanation of results
that help users to grasp and understand the space of retrieved
information using textual and graph modes.
8. Ranking of the results are not provided by Aemoo, Linked Jazz and
inWalk since the results are in the graph form.

Computing Semantic Similarity
1. SWOC and LED integrate DBpediaRanker into their system.
2. Yovisto uses a set of heuristics to determine the most relevant resources
related with a given DBpedia entity
3. Aemoo uses EKP which specifies the relevant types of resources and
defines the typical classes used to explain entities of a certain type.
4. LDSD (Linked Data Semantic Distance ) and DBrec (Music
Recommendations Using DBpedia) algorithms are used by Seevl to
retrieve artist recommendations.
5. Discovery Hub is based on a semantic spreading activation algorithm
coupled to a graph sampling technique.
6. Linked Jazz uses Mapping tool to build Linked Jazz Name Directory
dataset which is refined by the Curator Tool.
7. InWalk system uses HCf+ clustering algorithm to build a high level graph
called inCloud.

Database technique
• Yovisto use Freebase Parallax to let users’ query find a set of entities then
find the videos for these entities.
• SWOC and LED use Storage which stores similarity values between the
pairs of resources and the “popularity” of each resource in a DBMS for the
purpose of an efficient retrieval at runtime.
• Aemoo uses Knowledge Pattern (KP) Repository Manager that is
responsible for the storage, indexing and fetching of KPs.
• A Scalable RDF/NoSQL storage powered by OpenLink Virtuoso has been
utilized by Seevl for scalability purposes.
• Discovery Hub uses Virtuoso for user accounts and follows.
• Linked Jazz created a directory of personal names of jazz artists paired with
URIs called Name Directory.
• InWalk repository is created by inWalk system to provide a high-level view
of relevant linked data.

References
• Berners-Lee, T. (2006, July 27). Linked Data - Design Issues. Retrieved January 9, 2015, from
http://www.w3.org/DesignIssues/LinkedData.html
• Berners-Lee, T., Hendler, J., & Lassila, O. (2001). The semantic web. Scientific American, 284(5), p28.
• Castano, S., Ferrara, A., & Montanelli, S. (2014). inWalk: Interactive and Thematic Walks inside the Web of Data. (pp.
628–631). Presented at the EDBT.
• Dimitrova, V., Lau, L., Thakker, D., Yang-Turner, F., & Despotakis, D. (2013). Exploring Exploratory Search: A User Study
with Linked Semantic Data. In Proceedings of the 2Nd International Workshop on Intelligent Exploration of Semantic Data
(pp. 2:1–2:8). NY, USA: ACM.
• Fensel, D. (2005). Spinning the Semantic Web: bringing the World Wide Web to its full potential. Mit Press.
• Ferrara, A., Genta, L., & Montanelli, S. (2013). Linked Data Classification: A Feature-based Approach. In Proceedings of
the Joint EDBT/ICDT 2013 Workshops (pp. 75–82). NY, USA: ACM.
• Heath, T., & Bizer, C. (2011). Linked Data: Evolving the Web into a Global Data Space (1st edition). Morgan & Claypool.
Retrieved from http://linkeddatabook.com/editions/1.0/#htoc8
• Hitzler, P., Krotzsch, M., & Rudolph, S. (2011). Foundations of semantic web technologies. CRC Press.
• Jiang, T. (2014). Exploratory Search: A Critical Analysis of the Theoretical Foundations, System Features, and Research
Trends. In Library and Information Sciences (pp. 79–103). Springer. Retrieved from
http://link.springer.com/chapter/10.1007/978-3-642-54812-3_7
• Klyne, G., Carroll, J. J., & McBride, B. (2004, February 10). Resource Description Framework (RDF): Concepts and Abstract
Syntax. Retrieved January 15, 2015, from http://goo.gl/8P7WyQ
• Le Hors, A., Nally, M., & Speicher, S. (2012). Using read/write Linked Data for Application Integration-Towards a Linked
Data Basic Profile. Presented at the LDOW.
• Madhu, G., Govardhan, D. A., & Rajinikanth, D. T. (2011). Intelligent Semantic Web Search Engines: A Brief Survey. arXiv
Preprint arXiv:1102.0831. Retrieved from http://arxiv.org/abs/1102.0831

References Cont’d
• Marchionini, G. (2006). Exploratory Search: From Finding to Understanding. Commun. ACM, 49(4), 41–46.
http://doi.org/10.1145/1121949.1121979
• Marie, N., & Gandon, F. (2014). Survey of linked data based exploration systems. Presented at the IESD 2014-Intelligent
Exploitation of Semantic Data.
• Marie, N., Gandon, F., Ribière, M., & Rodio, F. (2013). Discovery Hub: On-the-fly Linked Data Exploratory Search. In
Proceedings of the 9th International Conference on Semantic Systems (pp. 17–24). New York, NY, USA: ACM.
http://doi.org/10.1145/2506182.2506185
• Mirizzia, R., Di Noiaa, A. R. T., & Di Sciascioa, E. Lookup, Explore, Discover: how DBpedia can improve your Web search.
• Mirizzi, R., Ragone, A., Noia, T. D., & Sciascio, E. D. (2010). Semantic Wonder Cloud: Exploratory Search in DBpedia. In F.
Daniel & F. M. Facca (Eds.), Current Trends in Web Engineering (pp. 138–149). Springer Berlin Heidelberg.
• Musetti, A., Nuzzolese, A. G., Draicchio, F., Presutti, V., Blomqvist, E., Gangemi, A., & Ciancarini, P. (2012). Aemoo:
Exploratory search based on knowledge patterns over the semantic web. Semantic Web Challenge.
• Passant, A. (2012). seevl: mining music connections to bring context, search and discovery to the music you like. From
challenge.semanticweb.org/submissions2011/swc2011_submission_11.pdf
• Pattuelli, M. C., Miller, M., Lange, L., Fitzell, S., & Li-Madeo, C. The Code4Lib Journal – Crafting Linked Open Data for
Cultural Heritage: Mapping and Curation Tools for the Linked Jazz Project [Journal]. Retrieved February 5, 2015, from
http://journal.code4lib.org/articles/8670
• Jenice Aroma R., & Kurian, M. (2013). A Survey on Tools essential for Semantic web Research. International Journal of
Computer Applications, 62(9), 26–29.
• Waitelonis, J., & Sack, H. (2012). Towards Exploratory Video Search Using Linked Data. Multimedia Tools Appl., 59(2),
645–672. http://doi.org/10.1007/s11042-011-0733-1
• White, R. W., Kules, B., & Bederson, B. (2005). Exploratory search interfaces: categorization, clustering and beyond:
report on the XSI 2005 workshop at the Human-Computer Interaction Laboratory, University of Maryland (Vol. 39, pp.
52–56). Presented at the ACM SIGIR Forum, ACM

THE END
Thanks for your attention

A Survey of Exploratory Search Systems Based on LOD Resources

More Related Content

What's hot (14)

Viewers also liked (12)

Similar to A Survey of Exploratory Search Systems Based on LOD Resources (20)

More from Karwan Jacksi (7)

Recently uploaded (20)

A Survey of Exploratory Search Systems Based on LOD Resources

Editor's Notes