ajitn
Joined: 20 Jul 2010 Posts: 6 Location: Chennai, India
|
Posted: Thu Aug 05, 2010 10:33 am Post subject: Relationship with the UNDL dataset |
|
|
Hi,
I am a little confused about the relationship between the activities of the UNLWeb (this site) and the UNL Center (www.undl.org).
The UNDL Foundation has made available a number of resources, including an enconverter and a deconverter framework, and a dictionary creator, to members of the UNL Society. However, I don't think the deconverter framework provided by the UNDL Foundation is compatible at all with the resources being developed on this site.
For example, the UNDL grammar rules look like this:
: "[a],ART" {N,^PRON,^TIME,^@pl,^VOW,^art:art} P100;
: "[an],ART" {N,^PRON,^TIME,^@pl,VOW,^art:art} P100;
which are totally different from the UNLWeb grammar rules, which look like this:
agt(%01,V;%02,N):=VS(%01;%02,+NOM);
and(%01;%02):=XA(%02,BEF;CC([and];%01));
My question is: will the resources on the UNLWeb interoperate with the 'official' resources on the UNDL website?
If not, is there a deconverter/enconverter being developed as part of the UNLweb for which these resources are being created?
-Ajit |
|
martins Site Admin
Joined: 16 Dec 2009 Posts: 1481 Location: Geneva, Switzerland
|
Posted: Fri Aug 06, 2010 1:08 pm Post subject: RES: Relationship with the UNDL dataset |
|
|
It’s important first to indicate that the UNLweb congregates now the “official” release of the UNDL Foundation’s tools. These tools (EUGENE, IAN, SEAN, NORMA, UNL Editor, etc) are available at the UNLdev, comply with the new UNDL Foundation’s guidelines, and are released under a GPL license. Unfortunately, this is not the case of the tools provided by the UNL Center in Tokyo, and that’s why the UNDL Foundation decided to discontinue EnCo, DeCo and DicBuilder, whose latest version is as old as 2006. The differences, however, concern not only copyright issues, but also technical aspects, which are detailed in the UNLwiki (Dictionary Specs and Grammar Specs). For a full picture of the UNDL Foundation’s new tools, users should refer to the road map available at http://www.unlweb.net/unlweb/index.php?option=com_content&view=article&id=65:road-map&catid=1:latest-news&Itemid=60.
As indicated in the Dictionary Specs, the new dictionary syntax is compatible, in several aspects, with the syntax of the old tools, but it has several new features that are not supported by the DicBuilder. They are signaled with an asterisk at http://www.unlweb.net/wiki/index.php/Dictionary_Specs. The grammar approach, however, is totally different: EnCo goes directly from lists to graphs, and DeCo from graphs to lists. The whole process involves rather low-level rules for manipulating generation and analysis windows. The new approach, which is more linguist-friendly, includes an intermediate step (the tree level), and is not bound to the sentence as the unit of analysis. That’s why there is no possibility of interoperation (at the grammar level) between the old and the new tools.
|
|