Results - Details
Search command : Author="Πλεξουσάκης"
And Author="Δημήτρης"
Current Record: 20 of 88
|
Identifier |
000428731 |
Title |
Keyword search over RDF using document-centric information retrieval systems |
Alternative Title |
Αναζήτηση μέσω λέξεων-κλειδιών επί RDF δεδομένων χρησιμοποιώντας εγγραφοκεντρικά συστήματα ανάκτησης πληροφοριών |
Author
|
Καντηλιεράκης, Γιώργος Δ.
|
Thesis advisor
|
Τζίτζικας, Γιάννης
|
Reviewer
|
Πλεξουσάκης, Δημήτρης
Φλουρής, Γιώργος
|
Abstract |
There are thousands of datasets published according to the principles of Linked Data and
Semantic Web. Many of those datasets, organized in RDF, are maintained either in crossdomain Knowledge Bases (e.g. DBpedia, Wikidata) or domain specific repositories (e.g.
DrugBank, MarineTLO), and are mainly used through navigation and structured query
languages like SPARQL. However these techniques are complex, lack flexibility and
possibly require a full knowledge of the underlying ontology. As a result, these datasets
are exploited by expert users only.
On the other hand, keyword search is the most widely used method for searching.
Keyword search is user friendly, offers instant content access, and keyword queries
support a wide range of expression while being extremely flexible. Information Retrieval
systems are designed for performing efficient keyword search in large data of
information, usually organized as full text documents. There are various highly
performant and effective state of the art search engines readily available. Such a search
engine is Elasticsearch, a distributed full text search engine that provides scalable search
over any kind of textual information.
In this thesis we introduce an approach for keyword-search over RDF datasets, by
adapting traditional IR techniques for both indexing and retrieval. Specifically, we test
how a dominant IR engine such as Elasticsearch, can be adapted for indexing RDF data
and enable keyword search. We provide a systematic analysis of different approaches to
cope with the challenges of indexing and retrieving structured information and exploiting
the graph capabilities of RDF. The response of the system comprises ranked RDF triples.
We also provide policies for ranking the different entities that are contained in these
triples, in order to support the requirements of entity search.
We report evaluation results of the different approaches in terms of: (i) the efficiency of
indexing and retrieval and (ii) the quality of retrieval. We test the effectiveness of our
system by evaluating the relevance of the constructed entities against the DBpedia-Entity
test collection, designed for entity search over the DBpedia KB and compare our results
to various state of the art systems. Our results showcase the effectiveness of the
proposed user friendly approach , that exploits the powerful features of scalable state of
the art search engines, and can be applied in any RDF dataset, with no prior knowledge
of the domain. The results show that Elasticsearch can effectively support keyword search
over RDF data, offering effectiveness comparable to that of systems built from scratch for
the task per se, that use entity-oriented and dataset-specic index structures.
|
Language |
English |
Issue date |
2020-03-27 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/e/a/b/metadata-dlib-1583220697-828481-12811.tkl
|
Views |
1017 |