The Cover PagesThe OASIS Cover Pages: The Online Resource for Markup Language Technologies
SEARCH | ABOUT | INDEX | NEWS | CORE STANDARDS | TECHNOLOGY REPORTS | EVENTS | LIBRARY
SEARCH
Advanced Search
ABOUT
Site Map
CP RSS Channel
Contact Us
Sponsoring CP
About Our Sponsors

NEWS
Cover Stories
Articles & Papers
Press Releases

CORE STANDARDS
XML
SGML
Schemas
XSL/XSLT/XPath
XLink
XML Query
CSS
SVG

TECHNOLOGY REPORTS
XML Applications
General Apps
Government Apps
Academic Apps

EVENTS
LIBRARY
Introductions
FAQs
Bibliography
Technology and Society
Semantics
Tech Topics
Software
Related Standards
Historic
Last modified: August 02, 2000
Dialogue Moves Markup Language (DMML)

[August 02, 2000] An article published in Communications of the ACM (CACM) presents the "Dialogue Moves Markup Language (DMML)" in the context of this CACM special issue on 'Personalization'. Excerpt: "The pragmatic goal of natural language (NL) and multimodal interfaces (speech recognition, keyboard entry, pointing, among others) is to enable ease-of use for users/customers in performing more sophisticated human-computer interactions (HCI). NL research attempts to define extensive discourse models that in turn provide improved models of context-enabling HCI and personalization. Customers have the initiative to express their interests, wishes, or queries directly and naturally, by speaking, typing, and pointing. The computer system provides intelligent answers or asks relevant questions because it has a model of the domain and a user model. The business goal of such compterized systems is to create the marketplace of one. In essence, improved discourse models can enable better one-to-one context for each individual. Even though we [IBM] build NL systems, we realize this goal cannot be achieved due to limitations of science, technology, business knowledge, and programming environments. [...] We would add another problem to this list: Our repositories of knowledge are not designed for NL interaction.. . . We address the issue of managing the business complexities of dialogue systems (for example, using NL dialogue on devices with differing bandwidth) by describing a piece of middleware called Dialogue Moves Markup Language (DMML). One of the key aspects of engineering is to design the middleware (middle layer)... which typically contains a dialogue manager, an action manager, a language-understanding module, and a layer for conveying messages between them... Universal Interaction uses the same mechanism for different communication devices (such as phones, PDAs, and desktop computers). This means conveying the same content through different channels by suitably modifying the way it is represented. Universal Interaction architecture is tailor-made for personalization: the actual interface for each user can be specifically constructed for him or her based upon geography-specific, user-specific, and style-specific transformations. How do we transform the content to fit into different representations? A potentially good idea is to use an XML/XSL-based architecture for the separation of form, content, style, and interactions. To make the idea more specific, imaging how one can represent a dialogue move in a stock trading scenario. DMML -- inspired by the theory of speech acts and XML -- is an attempt to capture the intent of communicative agents in the context of NL dialogue management. The idea is to codify dialogue moves such as greetings, warnings, reminders, thanks, notifications, clarifications, or confirmations in a set of tags connected at runtime with NL understanding modules, which allow us to describe participants' behaviors in terms of dialogie moves, without worring about how they are expressed in language. For instance, 'any positive identification must be followed by a confirmation.' The tags can also encode other parameters of the dialogue, such as the type of channel and personal characteristics. Thus, the dialogue can reflect the channel characterstics, which connects DMML and Universal Interaction. [Figure 2 presents a stock-trading transaction in DMML markup notation.] DMML thus illustrates the concept of communication middleware in dialogue systems, and is very well suited for personalization." See the full presentation: "Natural Language Dialogue for Personalized Interaction," by Wlodek Zadrozny, Malgorzata Budzikowska, J. Chai, Nanda Kambhatla, Sylvie Levesque, and Nicolas Nicolov. In Communications of the ACM (CACM) Volume 43, Number 8 (August, 2000), pages 116-120. Further background to this research has been published in "DSML [Dialog System Markup Language]: A Proposal for XML Standards for Messaging Between Components of a Natural Language Dialog System."

"Conversational Machines: Current web (and telephone) interfaces are menu driven. Thus, they are painful to use, because they organize information according to the enterprise offerings (merchandise, bank offerings, technical support information, etc.), and such organization rarely matches user's goals. As a result, accessing relevant information (e.g., a piece of support information) or performing a transaction (buying insurance) requires traversing several layers of menus, and often ends in failure. However, it is possible to allow users to express their desires directly, in natural language (augmented with pointing), and resolve goal ambiguities in a two or three steps of dialog. Our current project, Conversational Machines for the Web, explores natural language dialog interfaces for navigation, resolving customer problems and enhancing e-commerce by providing direct access to merchandise and data..."

'Wlodek Zadrozny [prior to 2000-08-02] managed the Conversational Machines project at IBM T.J.Watson Research Center. Currently, the project aims at identifying key technologies for building Web-based dialog interfaces for electronic commerce and customer support. Previously, the Conversational Machine group has built and evaluated with customers a number of successful prototypes of conversational telephony systems for telephony banking and directory assistance.']

References:

  • IBM Research - Conversational Machines

  • "DMML: An XML Language for Interacting with Multi-modal Dialog Systems." By Nanda Kambhatla, Malgorzata Budzikowska, Sylvie Levesque, Nicolas Nicolov, Wlodek Zadrozny, Charles Wiecha and Julie MacNaught (IBM T.J. Watson Research Center, US). Paper delivered at the Third International Workshop on Human Computer Conversation, Bellagio, Italy (3-5 July 2000).

  • "Natural Language Dialogue for Personalized Interaction." By Wlodek Zadrozny, Malgorzata Budzikowska, J. Chai, Nanda Kambhatla, Sylvie Levesque, and Nicolas Nicolov. In Communications of the ACM (CACM) Volume 43, Number 8 (August, 2000), pages 116-120. ['Technologies that successfully recognize and react to spoken or typed words are key to true personalization. Front- and back-end systems must respond in accord, and one solution may be found somewhere in the middle(ware).' The article presents the "Dialogue Moves Markup Language (DMML)" in the context of this CACM special issue on 'Personalization'.] "We address the issue of managing the complexities of dialogue systems (for example, using NL dialogue on devices with differing bandwidth) by describing a piece of middleware called Dialogue Moves Markup Language (DMML). One of the key aspects of engineering is to design the middleware (middle layer)... which typically contains a dialogie manager, an action manager, a language-understanding module, and a layer for conveying messages between them... Universal Interaction uses the same mechanism for different communication devices (such as phones, PDAs, and desktop computers). This means conveying the same content through different channels by suitably modifying the way it is represented. Universal Interaction architecture is tailor-made for personalization; the actual interface for each user can be specifically constructed for him or her based upon geography-specific, user-specific, and style-specific transformations. How do we transform the content to fit into different representations? A potentially good idea is to use an XML/XSL-based architecture for the separation of form, content, style, and interactions. To make the idea more specific, imaging how one can represent a dialogue move in a stock trading scenario. DMML -- inspired by the theory of speech acts and XML -- is an attempt to capture the intent of communicative agents in the context of NL dialogue management. The idea is to codify dialogue moves such as greetings, warnings, reminders, thanks, notifications, clarifications, or confirmations in a set of tags connected at runtime with NL understanding modules, which allow us to describe participants' behaviors in terms of dialogie moves, without worring about how they are expressed in language. For instance, 'any positive identification must be followed by a confirmation.' The tags can also encode other parameters of the dialogie, such as the type of channel and personal characteristics. Thus, the dialogue can reflect the channel characterstics, which connects DMML and Universal Interaction. [...] DMML thus illustrates the concept of communication middleware in dialogue systems, and is very well suited for personalization."

  • [August 02, 2000] "DSML [Dialog System Markup Language]: A Proposal for XML Standards for Messaging Between Components of a Natural Language Dialog System." By Dragomir R. Radev, Nanda Kambhatla, Yiming Ye, Catherine Wolf, and Wlodek Zadrozny (IBM TJ Watson Research Center, 30 Saw Mill River Road, Hawthorne, NY 10532. Email: {radev,nanda,yiming,cwolf,wlodz}@watson.ibm.com). In Proceedings of the Artificial Intelligence and Simulation of Behaviour (AISB) 1999 Convention; the workshop on Reference Architectures and Data Standards for Natural Language Processing, Edinburgh, UK, April, 1999. Abstract: "In this paper, we propose using standard XML messaging interfaces between components of natural language dialog systems. We describe a stock trading and information access system, where XML is used to encode speech acts, transactions and retrieved information in messages between system components. We use an XML/XSL based unification approach to display personalized, multi-modal system responses. We are proposing the creation of XML standards for all messaging between components of NLP systems. We hope that the use of XML in messaging will promote greater interoperability of both data and code. The working name of the proposed standard messaging language(s) is DSML ('Dialog System Markup Language.') [...] In our natural language dialog (NLD) processing system, a user can express requests in any modality (e.g., speech, text, graphics, etc.). The NLD system iteratively identifies the communicative act(s) of a user, queries for and fills the parameters of the identified action, executes the action, and displays the results to the user using an appropriate modality. Our class of NLD systems can be considered to be a sub-class of more general language engineering systems such as the General Architecture for Text Engineering (GATE) (Gaizauskas et al.) system. . . Conclusion and Future Work: We are proposing the use of XML/XSL technologies as a standard messaging format for the components of natural language dialog systems. The XML can be used at several levels. For instance, we can use XML for representing the logical forms output by NLP parsers and also to represent the dialog acts for dialog processing. Similarly, XML can be used to represent contextual information needed for utterance interpretation (e.g., background knowledge, previous discourse, etc.). We envision a phased approach to building XML based standards for NLP systems: Phase I There is a broad usage of XML for NLP/dialog processing systems leading to more modular and easily exchangeable code and data. However, each group uses their own custom XML. The groups communicate with each other. Phase II Standard XML languages are identified for different domains (e.g., brokerage systems, ATIS, directory services, etc.). Thus, we agree both on the syntax and the vocabulary of the representations. We foresee potential development of transducers to transform messages in other languages such as KQML to XML. Like XML, our XML-based standard will facilitate the encoding of speech acts such as 'request', 'assert', 'reply', etc. Phase III Standard semantics are identified for interpreting the standard XML languages, leading to interoperable data and code. This stage will involve the use of content languages such as KIF or Lisp, as well as standardized domain ontologies." See now 'DMML' (Dialogue Moves Markup Language). [cache]

  • Contact: Nanda Kambhatla, Conversational Machine Group manager. Or: Wlodek Zadrozny, IBM T.J. Watson Research Center, P.O. BOX 704, Yorktown Heights, NY 10598. Tel. +1 (914) 784 7835; Email: WLODZ@US.IBM.COM.


Hosted By
OASIS - Organization for the Advancement of Structured Information Standards

Sponsored By

IBM Corporation
ISIS Papyrus
Microsoft Corporation
Oracle Corporation

Primeton

XML Daily Newslink
Receive daily news updates from Managing Editor, Robin Cover.

 Newsletter Subscription
 Newsletter Archives
Globe Image

Document URI: http://xml.coverpages.org/dmml.html  —  Legal stuff
Robin Cover, Editor: robin@oasis-open.org