Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner: [368] trunk/src/dbpedia-navigator
|
From: <jen...@us...> - 2008-01-13 10:22:21
|
Revision: 368
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=368&view=rev
Author: jenslehmann
Date: 2008-01-13 02:22:16 -0800 (Sun, 13 Jan 2008)
Log Message:
-----------
extended ToDo list and minor improvements
Modified Paths:
--------------
trunk/src/dbpedia-navigator/ajaxfunctions.php
trunk/src/dbpedia-navigator/default.css
trunk/src/dbpedia-navigator/index.php
Modified: trunk/src/dbpedia-navigator/ajaxfunctions.php
===================================================================
--- trunk/src/dbpedia-navigator/ajaxfunctions.php 2008-01-11 17:01:00 UTC (rev 367)
+++ trunk/src/dbpedia-navigator/ajaxfunctions.php 2008-01-13 10:22:16 UTC (rev 368)
@@ -83,10 +83,19 @@
if(isset($triples['http://www.w3.org/1999/02/22-rdf-syntax-ns#type']))
$content .= '<p>classes: '.formatClassArray($triples['http://www.w3.org/1999/02/22-rdf-syntax-ns#type']).'</p>';
- // filter out uninteresting properties
+ if(isset($triples['http://dbpedia.org/property/reference'])) {
+ $content .= '<p>references: <ul>';
+ foreach($triples['http://dbpedia.org/property/reference'] as $reference)
+ $content .= '<li><a href="'.$reference.'">'.$reference.'</a></li>';
+ $content .= '</ul></p>';
+ }
+
+ // filter out uninteresting properties and properties which
+ // have already been displayed
unset($triples['http://xmlns.com/foaf/0.1/page']);
unset($triples['http://xmlns.com/foaf/0.1/depiction']);
- unset($triples['http://dbpedia.org/property/abstract']);
+ unset($triples['http://dbpedia.org/property/abstract']);
+ unset($triples['http://www.w3.org/1999/02/22-rdf-syntax-ns#type']);
// display the remaining properties as list which can be used for further navigation
Modified: trunk/src/dbpedia-navigator/default.css
===================================================================
--- trunk/src/dbpedia-navigator/default.css 2008-01-11 17:01:00 UTC (rev 367)
+++ trunk/src/dbpedia-navigator/default.css 2008-01-13 10:22:16 UTC (rev 368)
@@ -374,4 +374,9 @@
#footer {
clear: both;
text-align: center;
+}
+
+#todo {
+ padding: 15px;
+ margin: 10px;
}
\ No newline at end of file
Modified: trunk/src/dbpedia-navigator/index.php
===================================================================
--- trunk/src/dbpedia-navigator/index.php 2008-01-11 17:01:00 UTC (rev 367)
+++ trunk/src/dbpedia-navigator/index.php 2008-01-13 10:22:16 UTC (rev 368)
@@ -182,19 +182,64 @@
echo '><img src="images/valid-css.png" alt="valid CSS" /></a></div>'."\n";
?>
</div>
- <p><a href='rebuild.php'>rebuild [restart session and redownload WSDL file (for debugging)]</a></p>
- <p>
-ToDo:
-<ul>
- <li>get learning process working (with new SPARQL component)</li>
- <li>get local DBpedia SPARQL endpoint working</li>
- <li>many queries work correctly on the server, but yield to response in the interface (seems to be rather random)</li>
- <li>fix sometimes occurring PHP errors and warnings</li>
- <li>would be interesting to somehow view the Wikipedia article (without the left navigation part,
+ <p><a href='rebuild.php'>rebuild [restart session and redownload WSDL file (for debugging)]</a></p>
+</div>
+
+<div id="todo">
+<b>ToDo:</b>
+<ul style="float:left">
+ <li>Get learning process working (with new SPARQL component).</li>
+ <li>Get local DBpedia SPARQL endpoint working (next DBpedia release expected at the endof January and then every
+ two months, so it would be nice to have a script based partly automated or at least documented solution for
+ creating a DBpedia mirror).</li>
+ <li>Improve stability: Fix sometimes occurring PHP errors and warnings (check PHP error log).</li>
+ <li>Currently there are three SPARQL query methods: for labels, subjects, and article. These should
+ probably be replaced by a single web service method which allows to execute a SPARQL query. This will
+ greatly simplify the DL-Learner code while making the PHP code only slightly more complex. Such a method
+ also makes much more sense as a web service method, because other applications will need different SPARQL
+ queries. Similar to the learn method, there should be a threaded and a non-threaded version of the SPARQL
+ query method. This allows not to have the overhead of creating a separate thread for simple SPARQL queries
+ (especially since we assume that the DBpedia SPARQL endpoint is local).</li>
+ <li>Automatically learn concepts whenever an example has been added (and there is at least one
+ positive example present).</li>
+ <li>Show n (for example 5) instead of 1 best concept.</li>
+ <li>For each result, display a "+" which shows more information about the concept in an overlay box, e.g. its
+ Description Logic or OWL syntax, its classification accuracy on the examples, and which
+ examples it classifies (in-)correctly.</li>
+ <li>Move the "Learned Concepts" box above the main box in the center.</li>
+ <li>Remove the "Subjects from Concept" box and instead change the learned concepts to links (clicking
+ on a link shows instances of the concept).</li>
+ <li>Create a small number of test cases (e.g. 3), which can be used to verify that DBpedia Navigator is
+ working in typical scenarios (in particular cases where concepts with length greater one are learned).</li>
+ <li>Display "server call" in progress (or "n server calls in progress") in the top right corner
+ of the screen whenever AJAX queries are executed.</li>
+ <li>Allow to disable caching functionality (in Settings.php).</li>
+ <li>Fix the rebuild.php script such that PHP replaces the cached WSDL file by the new one.</li>
+ <li>Make DBpedia Navigator RESTful, e.g. URLs $base/showArticle/$URL for displaying an article;
+ $base/search/$phrase for searching; $base/listInstances/$complexClass for listing the instances of
+ a learned. Maybe session variables (in particuar the selected positive and negative examples) can
+ also be given, e.g. $base/search/$phrase?positives=[$URL1,$URL2,$URL3]&negatives=[$URL4]. The supported
+ URI design should be briefly documented (e.g. on a dbpedia.org wiki page). A good URI design allows
+ easier external access (just give someone a link instead of saying exactly which actions have to be done to
+ get to a state), simplifies debugging the application, and may be of use for creating further
+ features.</li>
+ <li>Improve search functionality [we will probably get feedback from Georgi in February].</li>
+ <li>[maybe] Display a tag cloud similar to <a href="http://dbpedia.org/search/">DBpedia search</a>.</li>
+ <li>Get a nice DBpedia Navigator logo (preferrably in SVG format) [currently in contact with Matt, but not
+ sure he has time to help].</li>
+ <li>[maybe] Instead of only allowing a search as entry point to the application, also display
+ a navigatable class tree.</li>
+ <li>[if possible] When expensive SPARQL queries or learning problems have been posed, there should be
+ some way to abandon these if the user has already switched to doing something else. Example: The user
+ has added 3 positive and 1 negative examples. This is executed as a learning problem, but has no solution (so
+ DL-Learner would run forever unless we pose some internal time limit). The user adds another negative example a
+ second later, so instead of letting the previous learning problem run for a long time (and needing much resources),
+ it should be stopped by DBpedia Navigator.</li>
+ <li>[if possible] Find an easy way to validate HTML/JS in AJAX applications.</li>
+ <li>[maybe] Would be interesting to somehow view the Wikipedia article (without the left navigation part,
tabs etc.) as an overlay, because the Wikipedia article will almost always be a human-friendlier
- description of an object compared to the extracted one</li>
+ description of an object compared to the extracted one.</li>
</ul>
-</p>
</div>
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
