Skip to contents |
W3C Incubator Activity

Minutes for November 6, 2007 CWL Meeting at TPAC2007

Attendees:

Agenda:

  1. CWL Architecture
  2. Top Ontology of CDL.nl
  3. Conversion from UNL to CDL and RDF
  4. OWL representation of UNL Knowledge Base
  5. Evaluation Methodology and Platform of CWL
  6. Extension of CWL-XG for 6 months
  7. Future Plan of CWL
  8. Table of Contents of XG Report

Minutes:

Introduction

1. CWL Architecture

Hiroshi: ``CWL: A Common Web Language for Humans and Computers'', presentation done at Yerevan, Armenia, 25 Sep 2007.
(1) Problems to be solved
Almost all Web pages are written in English. Google and others provide machine translation mechanisms. It is too much. It automatically translates. The translation is not in a good quality. No translation has 50% correctness. 50% is necessary to understand the meaning.
(2) Machine Understandability
HTML tag information is insufficient. RDF and OWL have no standard vocabulary.
(3) Objectives of CWL
CWL is for exchanging information and for computers to process information. It is to describe contents and meta-data of Web pages.
(4) Requirements for CWL
It needs to be independent from any natural languages. It should be able to be converted to/from each natural language. It needs to be a formal language. We can develop various kinds of controlled language from CWL. It needs to be able to be implemented in RDF/OWL.
(5) CWL Representation
CWL can be represented in UNL, CDL.nl or RDF/OWL XML. UNL is the language for communication. CDL.nl is the language for semantic computing. CDL.nl and RDF/OWL has no vocabularies, but UNL has. Our idea is to introduce UNL vocabularies to CDL.nl and RDF/OWL. UNL has a system which can convert to/from natural languages: French, Japanese, Chinese, ... more then 15 languages. CDL.nl has a mechanism of reasoning, so we can expect to combine them. CWL is not a language, but a model. We have CWL.UNL, CWL.CDL, CWL.RDF and so on.
(6) CDL
CDL is Concept Description Language, which is the language for semantic computing. CDL describes semantic/conceptual ...
(7) CDLs
CDL.core describes model and schema. CDL.nl is one type of CDL for natural language. CDL.unl means CDL which employs UNL vocabularies. CDL.jpn means CDL which employs Japanese vocabularies. Behind CDL, there are a lot of CDL languages.
(8) CDL.nl
CDL.nl consists of grammar, ontology and ...
(9) Natural Language Representation of CDL.nl
(10) Top Ontology of CDL.nl
The ontology is the same as one for UNL. We have nominal concept, predicative concept, 44 relations, attributes, ....
(11) UNL
UNL is an artificial language. A note represents a concept and an arc represents a relation. A node can be annotated by attributes.
(12) Relations
UNL has 46 relations.
(13) Attributes
(14) Types of UW
UNL has a formal way to define a concept.
(15) A kind of UW
UW expresses English concept.
`spring(icl>tool)' is the intersection of spring and tool. There are about 40 bounding relations.
(16) CWL representation
(17) UNL expression

Vahan: Demonstration Takes UNL document and illustrate the content of the document.
Views: Graph, UNL, CDL, RDF Automatically converted. There is a Web version available.

Martin: Is it similar to word net?

Vahan: UNL version is published and other languages can be deconverted from it. Each language has responsibility for deconversion. There is always a room for improvement for each language for better deconversion.

Nobuo: It is nice. You need better promotion.

Martin: Why don't you use RDF?

Vahan: All represent the same.

Hiroshi: UNL is invented in 1995 and at that time we did not have RDF.

Vahan: You can clearly see the advantage of UNL. It is compact. On the other hand, RDF can use various tools. Basically, you can use any format.

Nobuo: I heard Hebrew is the best language for representing semantics.

Hiroshi: The problem is that not a lot of people understand Hebrew.

Vahan: We need a precise translation, so we use a formal language. Any natural language is ambiguous.

Nobuo: Are those belong to united nations?

Hiroshi: All the property belong to united nations. We have three patents.

Matsuki: Is there any plan for UN to use?

Hiroshi: UNESCO is thinking to create Encyclopedia.

2. Top Ontology of CDL.nl

Hiroshi: The top of CDL.nl is concept: nominal concept, predicative concept, attributive concept, adverbial concept. Nominal concept has only one 'thing'. We have three types of predicative concept: do, occur and be. Top ontology itself is a big ontology because of the combination.

Tatsuya: Can you represent this in OWL?

Vahan: I can show an OWL version of UNL KB (http://www.undl.org). All the UNL KB can be represented in OWL.

Hiroshi: You can do the conversion automatically.

3. Conversion from UNL to CDL and RDF

Hiroshi: Already explained. Vahan has demonstrated it.

4. OWL representation of UNL Knowledge Base

Hiroshi: This has just demonstrated. OWL and UNL KB is very close.

5. Evaluation Methodology and Platform of CWL

Hiroshi: We can use the UNL system. It consists of deconverter and enconverter (convert to/from natural language). We can use the facility through UNL proxy. Our first motivation, the language barrier, can be overcome by this, but the second one, machine understandability, is not. We may use RDF tools.

Nobuo: What is KCIC?

Hiroshi: KCIC (Key Concept In Context) is an index to each UW in the UNL document. Using this, we can do any context search.

Hiroshi: We first thought to create CWL as a language, but finally it becomes an architecture with three languages under the architecture.

Nobuo: There are a lot of Web pages and Web users and sometimes metadata is provided. What happens if we have CWL?

Hiroshi: Metadata using English can be translated to CWL, then we can handle it.

Tatsuya: Can we have a even simpler version of UNL? Not for a natural language but just for expressing facts, tables, ...

Hiroshi: Even in that case, when we use English word, we have to specify the meaning, otherwise, machine cannot understand it.

Vahan: UNL for translation, CDL for semantic computing and RDF for navigation and aggregation.

Hiroshi: CWL editor (which helps to convert natural language to CWL) will be available by the end of December.

Nobuo: For general users, we would like to write Web pages easily.

Hiroshi: We will provide the editor.

6. Extension of CWL-XG for 6 months

Hiroshi: We need some more time to finish several applications and so on.

7. Future Plan of CWL

Hiroshi: At this moment, we are borrowing UW. We are proposing to develop a vocabulary for CWL. One problem of UW is that we cannot use UW for commercial use. If we define a new vocabulary, we can define UW as one of the language. I would like to propose this as Common Words.

Tatsuya: How long does it take?

Hiroshi: We are currently proposing to do it in three years.

Nobuo: Can you write the connection between current Web pages and Web pages written in CWL? Can you also explain where we need meta data or not? Please write those into the final report.

Nobuo: Please make a draft table of contents.

8. Table of Contents of XG Report

  1. Introduction (Uchida)
  2. CWL Architecture (Uchida)
  3. CWL.cdl Specification (Uchida)
  4. CWL.unl Specification (Uchida)
  5. CWL.rdf Specification (Vahan)
  6. Conversion among CWL.cdl, CWL.unl, CWL.rdf, UNLKB and OWL (Vahan)
  7. CWL system (platform)
    • CWL editor (Uchida)
    • UNL system (Uchida)
    • Semantic Computing (Yokoi)
    • RDF Applications (Vahan)
  8. Evaluation ()
  9. CWL and Web (Saito)
    • How CWL is useful for Web community.
  10. CWL in future (Ishizuka)
  11. Conclusion (Uchida)

Hiroshi Uchida, Common Web Language Incubator Group Chair
$Id: minutes20071106.html,v 1.22 2007/11/20 07:10:59 hagino Exp $
 Valid XHTML 1.0!