Query Construct Interfaces of RDF Data an introduction Qingxia Liu
Motivation Problem Functions in need Requirements How to get useful information in RDF dataset Functions in need Browse + Query Requirements Easy to use Easy to understand ( no need for training ) Portable (domain-independent) Support complex query
Approaches of Construct Structured Query Nature Language Input NLP-Reduce, Ginseng(CNL) Menu based visKWQL, Explorator, GRQL Form based K-Search Graph based Visor, Semantic-Crystal, Affective Graph, NITELIGHT Other VisiNav
Ginseng A guided input natural language search engine Advantages Base on a grammar Input full sentence by selecting the possible completions Advantages Easy to understand and use Disadvantages Too restrictive : cause to repeatedly reformulate in a kind of back tracking behavior
Ginseng
visKWQL A visual interface for the KWQL language to support constructing queries on semantic wiki Advantages Support complex query: conjunctions, disjunctions and negations of values or qualifier or resource terms; Allows for creation of rules: reshape and aggregate results Disadvantages Hard to read when there is too much layer
ci(text:Java OR (tag(name:XML) AND author:Mary)) visKWQL Select documents that either have “Java” in their text or that have the tag “XML” and were authored by Mary KWQL ci(text:Java OR (tag(name:XML) AND author:Mary)) visKWQL
Visor Overview-first data-on demand approach to browsing data Advantages Graphical representation Allows explore from multiple points in the graph Disadvantages Cannot construct complex query ( only support binary relations between collections)
Visor 在出生地拍过电影的演员
VisiNav Path traversal: select an object property to perform o set-based focus change Advantages Construct queries using navigation (object property) Easy to refine Disadvantages Only support star queries
VisiNav
Problems of querying RDF data Limited knowledge about target dataset exploratory: offering surrounding information at each step Difficult to use and understand Structured Query Language controlled natural language Structure ambiguous of NL e.g. “Find courses which are prerequisites of courses whose course department is computer Science and whose course credit is 3.”[1] tree-structured constructing procedure and represent
Our Method Interactive model Representation Construct structured query while exploring on schema level Representation Controlled Nature Language Avoid illegal query Structured (tree-like) lay out Easy to understand
Construct Procedure Init create the root node Modify Project Node Setting rename current node delete current node Value Constraint class constraint: for uri / blanknode/literal resource filter: value range binary constraint relation to existing variable Link add an arc to a new node Project
Controlled Nature Language Head: Find All (x) select all x that exists any patterns stated in the Query body. x is a list of projected variables. Body: that (statements) leading of the body part (x) has property… a button leading to recommends of available links (x) relate to … a button leading to recommends of available binary constraints (x) has (p y)/ (y) is p of (x) stating an added pattern (?x p ?y) x(C) C displays value constraints on variable x, and x is a button leading to node setting and value constraint
Example reconstruct: maintain an exploring tree Information Need: what do you want to find? restart reconstruct: maintain an exploring tree Find all ?film1 , ?runtime, ?director, ?writer, ?film2, ?runtime2, that ?film1 (Film) has runtime ?runtime (from 0 to 7200) has director ?director (Director) has made ?film2 (Film) has property… relate to… has writer ?writer (Writer) is writer of ?film1 has runtime ?runtime2 (from 7200 to 10800) Information Need: Find two films that has the same director and writer ,and the first film’s runtime between 0 and 7200 seconds, the second film’s runtime between 7200 and 10800 seconds;
References [1] Thompson, Craig W., et al. "Building Usable Menu-Based Natural Language Interfaces To Databases." VLDB. 1983. [2] A. Harth: VisiNav: A system for visual search and navigation on web data. Journal of Web Semantics, 8(4): 348-354, 2010 [3] E. Kaufmann and A. Bernstein. How useful are natural language interfaces to the semantic web for casual end-users? In 6th International Semantic Web Conference,pages 281–294, Nov 2007. [4] Nikolaos Athanasis, Vassilis Christophides, Dimitris Kotzinos: Generating On the Fly Queries for the Semantic Web: The ICS-FORTH Graphical RQL Interface (GRQL). International Semantic Web Conference 2004: 486-501 [5] Samur Araújo, Daniel Schwabe: Explorator: A tool for exploring RDF data through direct manipulation. LDOW 2009 [6] Khadija Elbedweihy, Stuart N. Wrigley, Fabio Ciravegna. (2012). Evaluating semantic search query approaches with expert and casual users. In The Semantic Web–ISWC 2012 (pp. 274-286). Springer Berlin Heidelberg. [7] I.Popov, M. Schraefel, W. Hall, N. Shadbolt: Connecting the Dots: A Multi-pivot Approach to Data Exploration. In: Proc. of International Semantic Web Conference, 553-568, 2011 [8] Andreas Hartl, Klara A. Weiand, Fran?ois Bry: visKWQL, a Visual Renderer For a Semantic Web Query Language. Proceedings of the 19th international conference on World wide web(pp. 1253-1256). ACM, 2010.