; SQR Semantic Query Rating Scheme
Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

SQR Semantic Query Rating Scheme

VIEWS: 1 PAGES: 1

  • pg 1
									SQR: A Semantic Query Rating Scheme
Hany Azzam, and Thomas Roelleke
{hany,thor}@eecs.qmul.ac.uk


1. Introduction                                                   4. Usage & Definitions
• A query rating scheme that identifies four lev-                  What is the rating of the query:
  els of rating for a semantic-expressive query                   “find titles of famous movies directed by Woody Allen where Woody Allen is a professor in the movie”?
  (semantic query in-short)
                                                                   Rating        Textual Description                    Propositions      Logical Representation
• The ratings reflect the interpretation assigned                  SQR-0          contexts containing the term           Φterm             retrieve(M) :-
  to a semantic query                                             words          woody and the term allen                                 M[woody & allen];
• The interpretations range from a traditional
  bag-of-words interpretation to more context-                    SQR-1          objects of type movie and display      Φattribute        retrieve(T) :-
  and semantic-aware interpretations                              structure      the attribute title                                      M.type(movie) & M.title(T);

                                                                  SQR-2          contexts directed by Woody Allen       Φclassification    retrieve(M) :-
                                                                  semantics      in which he is classified as a          Φrelationship     M.directedBy(woody_allen) &
                                                                                 professor                                                M[professor(woody_allen)];
2. Motivation
• A semantic-aware retrieval process impacts the                  SQR-3          famous movies                          Φvague            retrieve(M) :-
  query processing strategy                                       vagueness                                                               famousMovie(M);

• On a physical level, a semantic query usually
                                                                  SQR-123        titles of famous movies directed       Φattribute        retrieve(T) :-
  requires a search over different knowledge
                                                                                 by Woody Allen in which he is          Φclassification    famousMovie(M) &
  bases and indexes
                                                                                 classified as a professor               Φrelationship     M.directedBy(woody_allen) &
• On a logical level, a semantic query con-                                                                             Φvague            M.title(T) &
  tains words (terms) and classifications or                                                                                               M[professor(woody_allen)]];
  relationships
• How and which type of semantic informa-                                        The ‘?’ Denotes a Query and ‘&’ Denotes the Boolean ‘AND’ Operator
  tion has been considered when processing a
  query?                                                          Definition 1 SQR-0: Query q is an SQR-0 query iff each proposition ϕi ∈ q is a term proposition.

• What are the representation and processing                                                     q is SQR-0 : ⇐⇒ ∀ϕi ∈ q : ϕi ∈ Φterm
  strategies required to answer it?

                                                                  Definition 2 SQR-1: Query q is an SQR-1 query iff each proposition ϕi ∈ q is an attribute proposition.

                                                                                               q is SQR-1 : ⇐⇒ ∀ϕi ∈ q : ϕi ∈ Φattribute
3. Components
• The scheme relies on classifications, relation-
  ships, attributes and terms (words), which are                  Definition 3 SQR-2: Query q is an SQR-2 query iff each proposition ϕi ∈ q is either a classification
  all referred to as propositions                                 proposition or a relationship proposition.

• Classification: a class name with an object,                                          q is SQR-2 : ⇐⇒ ∀ϕi ∈ q : ϕi ∈ Φclassification ∪ Φrelationship
  e.g. “actor Woody Allen”
• Relationship: a subject with a relationship                     Definition 4 SQR-3: Query q is an SQR-3 query iff each proposition ϕi ∈ q is a vague proposition.
  name and object, e.g. “Mia Farrow worked
  with Woody Allen”                                                                              q is SQR-3 : ⇐⇒ ∀ϕi ∈ q : ϕi ∈ Φvague

• Attribute: an object with an attribute name
  and an atomic value, e.g. “The genre of Hus-
  bands and Wives is drama”                                       SQR-123 demonstrates one possible combination of the “basic” ratings. The rating contains the
                                                                  components of SQR-1, SQR-2 and SQR-3 and, hence, describes a query that has attributes, classifi-
• Term: a keyword, e.g. “allen”                                   cations and relationships.



References                                                        5. Benefits
[1] H. Bast, A. Chitea, F. M. Suchanek, and I. Weber. Ester:      • Provides the ability to communicate clearly and quickly the extent of semantics that is being inter-
    efficient search on text, entities, and relations. In SIGIR,
    pages 671–678. 2007.                                            preted for a query
[2] N. Fuhr, N. Gövert, and T. Rölleke. Dolores: A system
    for logic-based retrieval of multimedia objects. In SI-       • Predicts the suitable query processing strategy for each level of interpretation
    GIR, pages 257–265. 1998.
[3] G. Kasneci, F. M. Suchanek, G. Ifrim, M. Ramanath, and        • Provides a more “targeted” evaluation – more accurate assessment of a retrieval model’s effective-
    G. Weikum. Naga: Searching and ranking knowledge.               ness for each semantic query
    In ICDE, pages 953–962, 2008.
[4] J. Kim, X. Xue, and W. B. Croft. A probabilistic retrieval
    model for semistructured data. In ECIR, pages 228–239,
    2009.
[5] C. Meghini, F. Sebastiani, U. Straccia, and C. Thanos. A      6. Future Work
    model of information retrieval based on a terminologi-
    cal logic. In SIGIR, pages 298–308, 1994.                     • How to automate the query rating scheme?
[6] A. Trotman and B. Sigurbjörnsson. Narrowed extended
    xpath i (nexi). In INEX, pages 16–40, 2004.                   • How to associate the SQR with a hierarchy of retrieval models?
[7] R. van Zwol and T. van Loosbroek. Effective use of
    semantic structure in xml retrieval. In ECIR, pages 621–      • Do we need to define any aditional query ratings?
    628, 2007.

								
To top