Statement of Work

Our work will be conducted over a period of three years organized into 3 major categories of technology development (each with subcategories) and a final category for developing our application response to the challenge problems:

1. Collaborative and Interactive Acquisition of Knowledge
   • Acquiring Annotation Knowledge

   • Acquiring Ontologies, Partial Domain Theories, Lexicon

   • Acquiring Rule-like and Procedural Know-how.

2. Techniques for Acquiring Knowledge from Static Assets
   • Extracting Knowledge from Texts

   • Extracting Knowledge from Data Bases

3. Techniques in Support of Knowledge Acquisition
   • Managing Multiple Perspectives and Achieving Consensus

   • Incorporating Heterogeneous Assets using Protocol of Inference

   • Achieving Natural Interaction using the Intelligent Room

4. Challenge Problem Application Development

Work will be conducted in most tracks in each of the three years with the balance of effort dictated by the overall goals of the HPKB program. In particular, in the later years we will put more effort into building the ``application response'' to the challenge problems while in the early years, we will focus on enabling technology and core knowledge acquisition.

Our techniques should be relevant to any of the ISO programs mentioned in the BAA, but we feel that our focus on broad based, textual knowledge is particularly relevant to crisis understanding and intelligence management problems.

The anticipated focus of effort for each track is as follows:

1. Collaborative and Interactive Acquisition of Knowledge

   1. Acquiring Annotation Knowledge:

      1. In Year 1 we will:
         1. Precisely define the annotation framework
         2. Publish an annotators guide on the Web
         3. Create a Web-based interface capturing natural language annotations of documents
         4. Test the annotators interface through in-house use by sophisticated users
         5. Explore statistical techniques for topic sentence spotting

      2. In Year 2 we will:
         1. Revise the annotator's interface based on the experience with in-house use
         2. Test the updated annotator's interface in controlled use by a small set of domain experts
         3. Explore the use of linguistic clues for topic sentence spotting

      3. In Year 3 we will:
         1. Revise the annotator's interface based on the year 2 experience of the domain experts
         2. Deploy the annotator's interface for unrestricted use by a broad group of domain experts

   2. Acquiring Ontologies, Partial Domain Theories, Lexicon

      1. In Year 1 we will:
         1. Precisely define the framework for background knowledge capture
         2. Publish on the Web a guide for providers of background knowledge
         3. Create a Web-based natural language interface for entering ontological information
         4. Create a Web-based natural language interface for entering lexicon entries
         5. Test the annotators interface through in-house use by sophisticated users

      2. In Year 2 we will:
         1. Revise the interface based on the experience with in-house use
         2. Test the updated interface through controlled use by a small set of domain experts
         3. Explore the use of problem solving failures to guide acquisition of background knowledge

      3. In Year 3 we will:
         1. Revise the interface based on the Year 2 experience of the domain experts
         2. Deploy the interface for unrestricted use by a broad group of domain experts
         3. Deploy technology guiding acquisition of background knowledge by analysis of problem solving failures

   3. Acquiring rule-like and procedural know-how. (Rule-like information includes If-Then rules, influence diagrams, and the like.)

      1. In Year 1 we will:
         1. Precisely define the framework for capturing rule-like and procedural knowledge
         2. Publish on the Web a guide for providers of rule-like and procedural knowledge
         3. Create a Web-based natural language interface for acquisition of rule-like and procedural knowledge
         4. Test the interface through in-house use by sophisticated users
         5. Explore the Use of Intelligent Room technology for acquisition of rule-like and procedural knowledge

      2. In Year 2 we will:
         1. Revise the Web-based interface based on the experience with in-house use
         2. Test the updated Web-based interface through controlled use by a small set of domain experts
         3. Test the User of Intelligent Room technology with a small set of domain experts

      3. In Year 3 we will:
         1. Revise the Web-based interface based on the Year 2 experience of the domain experts
         2. Deploy the Web-based interface for unrestricted use by a broad group of domain experts

2. Techniques for Acquiring Knowledge from Static Assets

   1. Extracting knowledge from Texts

      1. In Year 1 we will:
         1. Identify a set of documents which exemplify the type of from which we wish to extract knowledge
         2. Test our syntactic coverage of these documents
         3. Improve our syntactic coverage
         4. Publish a document characterizing machine parseability

      2. In Year 2 we will:
         1. Devise techniques to operationalize the content of machine parseable documents
         2. Demonstrate the acquisition of partial domain theories from machine parseable documents
         3. Continue to improve our coverage

      3. In Year 3 we will:
         1. Demonstrate the acquisition of rule-like knowledge from machine parseable documents

   2. Extracting Knowledge from Data Bases

      1. In Year 1 we will:
         1. Extract time-stamped factual information from data bases
         2. Recognize changes and events from time-stamped factual information

      2. In Year 2 we will:
         1. Recognize changes and events at multiple temporal granularities

      3. In Year 3 we will:
         1. Demonstrate combined extraction of information from dissimilar data bases
         2. Relay information between dissimilar data bases

3. Techniques in Support of Knowledge Acquisition

   1. Managing multiple perspectives and achieving consensus

      1. In Year 1 we will:
         1. Extend our current techniques for representing multiple contexts of belief
         2. Develop tools to identify information as a ``near miss'' to current information
         3. Develop a Web-based interface for interactions leading to consensus
         4. Test the interfaces through in-house use by sophisticated users

      2. In Year 2 we will:
         1. Improve the interface based on experience with in-house testing
         2. Extend our Web-based interface to allow a user to situate his knowledge within a perspective
         3. Develop techniques for comparing the problem solving behavior emitted in different perspectives
         4. Incorporate these techniques into the interface
         5. Test the updated interface in controlled use by a small set of domain experts

      3. In Year 3 we will:
         1. Revise the interface based on the Year 2 experience of the domain experts
         2. Deploy the interface for unrestricted use by a broad group of domain experts

   2. Incorporating Heterogeneous Assets using Protocol of Inference

      1. In Year 1 we will:
         1. Work with other HPKB contributors to refine the hierarchy of key steps in the protocol of inference
         2. Publish the updated description of the protocol steps
         3. Demonstrate that we can incoporate existing ontologies and reasoners into our System using the Protocol

      2. In Year 2 we will:
         1. Demonstrate that we can use our System to capture knowledge for knowledge bases and reasoners external to our system

   3. Achieving Natural Interaction Using the Intelligent Room
      1. In Year 1 we will:
         1. Develop machine vision techniques for understanding box and arrow diagrams
         2. Develop techniques for coordinating visual diagram understanding with linguistic descriptions of the semantics of the diagram's components

      2. In Year 2 we will:
         1. Demonstrate the use of the technology developed in year 1 to capture procedures, plans, indicator and warning chains, etc.

      3. In Year 3 we will:
         Use this technology to develop our response to the Challenge problem

4. Challenge problem application development

   1. In Year 2 we will:
      1. Acquire/incorporate core ontologies
      2. Acquire domain know-how, procedures, rules, and facts
      3. Use the Intelligent Room for acquisition

   2. In Year 3 we will:
      1. Demonstrate the challenge problem application in Intelligent Room and over the Web.