In Response to Tasking for:

F-30602-00-0179

DARPA Agent Markup Language (DAML)

DAML for Services

 

Intent of Work

 

for

The Defense Advanced Research Projects Agency

April 13, 2001

Submitted by

Booz·Allen & Hamilton, Inc.

 

 

Technical Point  of Contact

Administrative Point of Contact

Mr. Paul Chi

Mr. Pete Vaeth

Booz·Allen & Hamilton, Inc.

Booz·Allen & Hamilton, Inc.

8283 Greensboro Drive

8283 Greensboro Drive

McLean, VA 22102

McLean, VA 22102

Phone:  (703) 902-6978

Phone:  (703) 902-6939

Fax:  (703) 902- 7117

Fax:  (703) 902-3650

Chi_paul@bah.com

vaeth_peter@bah.com

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



Intent of Work (IOW)

A.   Technical goals

Booz٠Allen proposes to develop tools focused in the following two areas:

·        XML to DAML translation

·        Query Relevance Assessment

In addition, Booz٠Allen will continue our close coordination with the Stanford Knowledge Systems Laboratory Research Team to collaboratively enhance research and development of the Annotation tool.  The following sections outline our technical goals, our collaboration partners, the intended user community, and our long term plans.

 

A.1             XML to DAML Translator: 

Answers to IOW Questions:

1. What is the technical goal/accomplishment you are hoping to achieve?

Booz·Allen proposes to develop a  DAML development tool that can translate XML schema and data sets into DAML schema and content respectively.  The approach is as follows:

1.      We will first work on  the translation of XML DTDs to DAML schemata.

2.      Next we will work on the translation of DTD-based XML content into DAML content.

3.      We will then work on the translation of XML schemata into DAML schemata, and finally

4.      We will translate XML-schema-based XML content into DAML content.

 

The tool is a functional module, intended for use by any user community including the current DAML development community that is focused on developing DAML-based applications. Initially the tool will be made available as sharable object to be bound into higher-level development tools.  The sharable object will be implemented as a Java Archive (JAR) file executable from the command line. A published API will be included, created by the javadoc tool.

 

The translator tool will then be made available as a family of DAML-enabled Web services via the Stanford DAML service authoring tool combined with Stanford's Web service axiomatization methodologies.

 

The current intended relationship between the XML-to-DAML translation tool and ontology translation/mapping tools is as follows:  The intent is to decouple the process of XML-to-DAML translation from the process of DAML ontology translation/mapping.  The two processes can be combined by piping the results of XML-to-DAML translation into DAML ontology translation/mapping.  And so each XML DTD gets its own new DAML schema.  This new DAML schema is then potentially mapped to existing (or future) DAML schemata.  Such mappings may include alias definitions.

2. Who are you grouping with in the DAML program and/or who is the intended military user?

We are grouping with Stanford in converting our sharable object into a Web service with their annotation tool. The XML to DAML tool will be a self-contained tool, providing the capability to all members of the DAML language development community to access and utilize. The tool could be used by anyone within the DAML development community who needs XML to DAML translation. Any military user could be a potential user of this tool. For example, Intelink is migrating from HTML to XML and this tool would be beneficial to easily DAML-ize their web sites.

3. Who is the intended user of what you are developing, and why would he/she use it (i.e. the "lifecycle" thing)?

The intended users for the tool are primarily the developers of DAML-based applications in commercial and government domains. As organizations increasingly adopt DAML, such a tool would be highly beneficial for knowledge managers to migrate their enormous amounts of XML content into DAML easily. The ease of conversion would in turn increase the acceptance of DAML. Since most web sites are migrating towards XML this tool will allow them to quickly convert their content to DAML without additional software development.

4. What do you think the next logical step will be in the 1-2 years after?

The next steps for the XML-DAML translator are to convert it into a service using the annotator tool.

 

Future steps for the XML-DAML translator:

 

1. Exploration of a DAML XSLT Implementation. XSLT is a powerful tool for processing and transforming XML documents.  We would like to extend the work of other teams in the areas of DAML XSLT and use BBN's work in this area (e.g., a form of XSLT tools, Apache xalan, etc.). Our goal is to offer the power of XSLT to DAML via a DAML XSLT or a DAML Document Object Model (DOM) that will allow processing / transformation of DAML documents or knowledge bases.  Our intention is to work towards a DAML XSLT toolkit.

 

2. Interface to the E-Speak Vocabulary (base ontology) and/or the BizTalk Framework.  Since the intention of the XML-DAML translator is to continue to translate , the logical next steps for the tool would be to begin work on gateways to e-commerce specifications such as E-Speak and BizTalk. One of the lower layers in the gateway involves translation from ESpeak vocabulary/BizTalk framework to DAML.

 

3.  HotDAML Wireless Component. The wireless arena is an important part of our ongoing research and development efforts. We are prepared to build upon the XML-to-DAML translation, as well as the XSLT research and collaborate with Stanford’s wireless research and development initiatives. Our approach for building a wireless component involves HotDAML Wireless creation through the utilization of the XML-to-DAML Translator, the Wireless Markup Language (WML), and the Wireless Access Protocol (WAP) communications protocol.  Our goal is to access DAML services via a wireless interface to DAML that is displayed on a specified handheld device.

 

Next steps beyond the development of these components potentially are:

 

-  Combine the XML-DAML translator and the DAML XSLT into one tool or plug-in to be utilized by other DAML tools.

-  Continue to extend DAML XSLT or DAML DOM

-  Continue research on a DAML wireless gateway – enabling WML-based application the ability to access DAML services.

 

 

A.2             Query relevance assessor

Answers to IOW Questions:

1. What is the technical goal/accomplishment you are hoping to achieve?

Booz Allen proposes to build a query relevance assessing tool to demonstrate how DAML will be used to enhance Web text retrieval. In particular, we will consider how the DAML language could be used to refine relevance rankings of Web page documents with respect to terms within search engine queries. The tool will assess relevance rank weights of DAML mark-up with respect to individual query terms.

 

Ontology classes and properties contain a tremendous wealth of contextual knowledge regarding relevance of Web pages (or Web-based text in general) to their document terms. DAML ontologies give rise to a weighted, hierarchical graph-theoretic structure with ontology classes, properties, and their instances as nodes. We will exploit these features of DAML ontologies to impart a powerful semantic component to term weight assignment for document ranking.

 

The DAML mark-up whose relevance our tool will assess, may reference only a single ontology or ontology class (or may in fact actually be a single ontology, ontology class, or ontology class instance) or perhaps would reference various different ones. Initially we will assume that a query term is actually a class or class instance within all the ontologies referenced by the mark-up. In the future, we will enhance the tool to process multi-term queries.

 

A particular thrust for this effort will be to show how this form of DAML-oriented term weight assignment may be accomplished when only a very limited amount of background information, if any, regarding the query (other than the user's actual query terms themselves of course) is imparted to the tool.

2. Who are you grouping with in the DAML program and/or who is the intended military user?

We intend to group with the JHU/APL group which is also concerned with the DAML implications for information retrieval. Part of their research involves finding and collecting classes and ontologies relevant to search queries but apparently they are not specifically confronting the issue of ranking these classes and ontologies with respect to their relevance to the query once they have been collected. It is this issue on which we at Booz Allen wish to concentrate now. We have discussed a partnership with JHU/APL with Dr. James Mayfield as well as Paul McNamee and they view the idea favorably. We will team with them to design a uniform API for our combined efforts.

 

As suggested above, the tool we propose is in particular a step toward DAML-enhanced Web text retrieval so the intended military user may be from any branch since all are interested in effective information retrieval. But, we believe the intelligence community is particularly keenly interested in this problem.

3. Who is the intended user of what you are developing, and why would he/she use it (i.e. the "lifecycle" thing)?

As indicated, the intended user is anyone interested in enhanced document retrieval since the tool could be used to help rank retrieved Web documents (or Web-based text in general) with respect to their relevance to a WWW query search. It could also be used to rank ontologies or ontology classes themselves with respect to a query and these could in turn then be used, for example, to add contextual information to significantly inform the search.

4. What do you think the next logical step will be in the 1-2 years after?

The next logical steps in the 1-2 years after should include extension and enhancement of our core tool. Such extension and  enhancement may include showing how the DAML-enhanced term weights computed by the tool may be applied to the design and construction of an extended Boolean information retrieval system for the DAML-based Web that will use relaxed Boolean operators to combine terms for multi-term queries. We also will demonstrate how our tool functions when used in tandem with various well-known text retrieval approaches such as Singular Value Decomposition (SVD) oriented techniques for Latent Semantic Indexing as well as when it is coupled with methods that enable it to exploit the citation link structure of  WWW pages.