Falcon-S - Semantic Search on Soccer Images


	Basic Search		Semantic Search		Explore		Help

Content

About Falcon-S
Semantic Query Definition
     Basic Concept
     Create New Semantic Query
     Edit Existing Semantic Query
Browser Compatibility

About Falcon-S

Falcon-S is a semantic web application focusing on soccer domain.

Powered by domain knowledge base and pre-defined basic semantic associations between the most importan classes (such as: soccer palyer, team, competition and etc.) , Falcon-S can provide more interesting and powerful searching and browsing mechanisms than traditional search engines can do.

At present, Falcon-S provides the following services:

Basic search. Falcon-S provides basic service for searching objects (e.g., players, teams), semantic associations among objects (e.g. soccerPlayer playFor Team) and images about identified objects in the soccer domain.
Semantic search. Falcon-S provides an advanced service for searching objects by semantic query driven by domain-specific ontology. It provides a list of semantic queries in common use, and more importantly, a graph-based user interface to let users customize their semantic queries based on existing ones or define new semantic queries from scratch.
Explore. Falcon-S provides a service for exploring objects in domain-specific ontology and knowledge base. Using Falcon-S, users can explore objects semantically associated by domain konwledge, and browse crawled images related with identified objects.

Falcon-S is based on ontology and related technologies. An ontology (download) and a knowledge base for the soccer domain are built by integrating the information from the official web site of FIFA, UEFA and existing work like [1] etc. It crawls the World Wide Web for images on soccer domain, analyses their context and indexes them on the individuals (e.g., players, teams) in the knowledge base. Keyword-based searches are mapped to one or more such individuals, semantic associations among objects are extracted with an ontology-driven approach, and the related images are returned. Objects can be semantically explored through an ontological view. Most importantly, Falcon-S provides an ontology-driven approach to let users define their own semantic queries for searching domain-specific objects. This novel feature is implemented by using some advanced technologies, such as Ajax, SVG and etc. In addtion, Falcon-S facilitates the low-level features of crawled images such as color histogram to improve the performance.

Falcon-S is a member of the Falcon family developed by the XObjects Group,
Southeast University, China.
Contact us: hhwu AT seu.edu.cn

Semantic Query Definition
(To view or edit semantic query, you may need the lastest version of SVG Viewer. Adobe Viewer is prefered.)

The graph drawed in the SVG Viewer represents a semantic query. The nodes are classes and the edges are semantic associations between these classes. Each node in the graph is colored. The legend part of sample query 1 illustrates the meaning of the colors.

Sample Query 1

Legend

Nodes need to be set constrains on or set as query variable.

Node that currently selected.

Node that set as query variable.

Black nodes.

Nodes that no more action are needed.

If both of the requirements listed below are met, a semantic query would be in its finish state. (Query and Save actions will be available.)

One and only one node is defined as query variable.
All leaf nodes (Nodes whose edges number equals one.) are defined. Either defined as query variable or set constrain on it by specifying the node as an instance of its type.

In one semantic query graph, there are three kinds nodes at most.

Variable Node. Node set as query variable.
Black Nodes. Nodes whose edges number is greater than one and also no constrains were defined on it.
Instanced Nodes. Nodes were set constrains on by specifying one instance of its type on it.

Create new semantic query

There are total 4 steps to create a new semantic query from scratch.

In the Semantic Search page, click the new button. The creating new semantic query panel would be shown.

Step 1. Select a class.	Select a class in the list, then associations related to this class will be shown in next step.
Step 2. Select an association.	Select an association from the list, then system will initialize a graph that represent your selected association.
Step 3. Edit query on the graph.	When selecting nodes in the graph, edit toolbar will be shown in th right side of the graph. For each selected node, three actions are available. set constrains on it by specifying one instance on it. set it as query variable. append associations on it. It is notable that you can select two nodes in the graph at the same time. This makes it possible to connecte two existed nodes with the possible associations between them. Deletion actions are not available now, but they would be implemented soon. If the graph was edited into its finish state (see the two requirements to be met above), query and save toolbar will be shown.
Step 4. Save or query the graph.	When in step 4, your query is in its finish state. Query and save toolbar has been shown(see the illustration below). You can save your query or run it. Save/Query toolbar shown in step 4.

Edit defined query

Select one of the defined queries in the query list. Coresponding semantic query graph will be shown. Click on nodes to edit it. The edit actions are the same as what is mentioned in the Create new query section.

Browser Compatibility

Falcon-S is powered by AJAX, SVG, etc. There are a lot of well known compability problems about these technologies across popular browsers, such as XMLHttpRequest, SVG support, etc. The support for mainstream browsers is listed as follows.

Browser	Services
Browser	Basic Search	Semantic Search	Explore
IE (With SVG Plugin)	Y	Y	Y
Firefox 1.5 or Higher	Y	Y	Y
Opera	Y	N	Y

Reference

[1] C. Tsinaraki, P. Polydoros, F. Kazasis and S. Christodoulakis, Ontology-Based Semantic Indexing for MPEG-7 and TV-Anytime Audiovisual Content, Multimedia Tools and Applications, 26 (2005), pp. 299-325.