MAin Body
Time POints and goals
MAIN

HITIQA Project Management Plan

Lead Organization: SUNY Albany (PI: Tomek Strzalkowski)

Subcontract: Rutgers University (co-PI: Paul Kantor)

POP: JAN02 ® DEC03

 

Task 1. Interactive Question Answering

Task Responsible: Strzalkowski (SUNY)

Start® End: JAN02 ® DEC03

Dependencies & Risks: 2; 3; 5

Goals & Deliverables: User interactive question answering system HITIQA

Description: Develop the core question answering system

Subtask 1.1. Question Analysis and Classification

Responsible: TS

Start® End: JAN02 ® DEC02

Dependencies & Risks: none

Goals & Deliverables: Improve efficacy of semantic processing and dialogue functions by eliminating unlikely interpretations of questions and providing guidance for conceptual clustering. Question Processing and Typing Module (QPTM).

Description: Develop methods and tools for question analysis. Develop or adapt question classification based on the results of question analysis and user performance data.

Step 1.1.1. Question Analysis

Responsible: TS

Start® End: JAN02 ® JUN02

Dependencies & Risks: None. QPM module is expected to improve semantic processing but is not critical.

Goals & Deliverables: Improved pre-semantic analysis of questions. Question Pre-processing Module (QPM)

Description: Identify key terms and phrases, key concepts, focus, main functional elements in the question.

Step 1.1.2. Question Typology

Responsible: TS

Start® End: JUL02 ® DEC02

Dependencies & Risks: 1.1.1

Goals & Deliverables: Improved pre-classification of questions. Question Typing Module (QTM)

Description: Adapt and expand question typology based on systems developed for TREC.

Subtask 1.2. Data-Driven Semantics for Questions

Responsible: TS

Start® End: JAN02 ® SEP02

Dependencies & Risks: 1.1

Goals & Deliverables: Semantic Interpretation Module for HITIQA system.

Description: Develop algorithms for subcategorizing the Retrieved Set. Experiment with granularity (passages, events); seeded clustering, dimensionality reduction.

Step 1.2.1. Attribute-Value Pairs

Responsible: TS, SS

Start® End: JAN02 ® JUN02

Dependencies & Risks: 1.1.1

Goals & Deliverables: Attribute structure of the questions

Description: This task will attempt to uncover the attribute structure of the question by locating attributes and/or their values. These attributes may belong to multiple relations thus leading to potentially ambiguous interpretations of the question.

Step 1.2.2. Text Chunking

Responsible: TS, SS

Start® End: JAN02 ® JUN02

Dependencies & Risks: 1.1.1

Goals & Deliverables: Document text divided into meaningful chunks

Description: This task will develop a module to chunk text into smaller units that will be assessed as building blocks for the information required to compose an answer to a question. The initial option is to use naturally occurring paragraphs.

Step 1.2.3. Topical Clustering

Responsible: TS, SS

Start® End: FEB02 ® SEP02

Dependencies & Risks: 1.1.1, 1.2.1, 1.2.2

Goals & Deliverables: Retrieved Set broken into polyhierarchies

Description: This task will construct a module to derive a hierarchical clustering of text chunks from "relevant" documents into clusters representing alternative and/or complementary interpretations of the questions.

Subtask 1.3. Integrate Clarification Dialogue

Responsible: TS

Start® End: OCT02 ® DEC03

Dependencies & Risks: 2

Goals & Deliverables: Clarification Dialogue Module for HITIQA system

Description: Integrate Clarification Dialogue Manager into Q&A main system.

Subtask 1.4. Answer Generation

Responsible: TS

Start® End: JAN02 ® DEC03

Dependencies & Risks: 1.2; 3; 5

Goals & Deliverables: Answer Generation module for HITIQA system

Description: Develop guidelines for producing final answer to the user. These include criteria for composition, length, order and layout.

Step 1.4.1. Cut-and-Paste Answers

Responsible: TS, TL

Start® End: JAN02 ® DEC02

Dependencies & Risks: 1.2.3

Goals & Deliverables: Answers composed of concatenated text fragments determined to be relevant to the question topic and of high quality.

Description: Topical clusters determined to be relevant and complementary will yield candidate passages.

Step 1.4.2. Integrate Quality Assessments

Responsible: TS, TL

Start® End: JAN02 ® DEC02

Dependencies & Risks: 1.2.3; 1.4.1; 3.4

Goals & Deliverables: Answers composed of text fragments judged for information quality determined to be relevant to the question topic and of high quality.

Description: Improve answer generation capabilities by including information quality assessments in the final selection of sources.

Step 1.4.2. Generative Answers

Responsible: TS, TL

Start® End: JUN03 ® DEC03

Dependencies & Risks: 5.3

Goals & Deliverables: Answers that are composed from fused and de-conflicted text fragments

Description: Replaces cut-and-paste answer generation with fused answers, which are expected to be more concise and better organized at the conceptual level. Depends upon a successful completion of answer fusion task (5.3).

Task 2. Clarification Dialogue

Task Responsible: Strzalkowski (SUNY)

Start® End: JUN02 ® DEC03

Dependencies & Risks:

Goals & Deliverables:

Description:

Subtask 2.1. Quantitative text processing

Responsible: TS

Start® End: JUN02 ® DEC02

Dependencies & Risks: 1.2.1

Goals & Deliverables: Text processing to support interactive dialogue and visualization

Description: Extend semantic processing of questions to support dialogue management. In particular, the Retrieved Set structure will provide dialogue motivators whenever a clarification of user request is necessary.

Step 2.1.1. Text Chunk Filtering

Responsible: TS, SS

Start® End: JUL02 ® JUN03

Dependencies & Risks: 1.2.1

Goals & Deliverables: Retain only "relevant" chunks

Description: This task improves text chunk clustering so that it can efficiently support human-computer dialogue. In particular, clusters need to be formed around concepts that are salient to the user and relevant to the question topic.

Step 2.1.2. Cluster Signatures

Responsible: TS, SS

Start® End: SEP02 ® JUN03

Dependencies & Risks: 1.2.3, 2.1.1

Goals & Deliverables: Cluster signature support differentiation between clusters. Each cluster will be assigned a unique signature describing its content.

Description: This task will develop algorithm to derive signatures for clusters in order to differentiate their content in terms of attributes and their values. This will in turn support clarification dialogue by generating dialogue motivators from signature differentials.

Subtask 2.2. Dialogue Management

Responsible: TS

Start® End: JUL02 ® JUN03

Dependencies & Risks: 1.2; 2.1

Goals & Deliverables: Dialogue Manager (DM) module controls and facilitates interaction between the user and the data.

Description: Develop Dialogue Manager to conduct dialogue with the user. The dialogue manager will be driven by the question semantics, user profile and context information, and answer quality constraints.

Step 2.2.1. Theme Router

Responsible: TS, SS

Start® End: JUL02 ® JUN03

Dependencies & Risks: 1.2.3

Goals & Deliverables: Cluster signature support differentiation between clusters. Each cluster will be assigned a unique signature describing its content. The goal is to have each cluster representing a possible interpretation of the question such that they are either mutually exclusive or complementary.

Description: This task will select likely interpretations for the question and adopt a polyhierarchy of candidate meanings. The Theme Router will rank possible interpretations and generate a dialogue move to ask the user for clarification whenever multiple possibilities exist.

Step 2.2.2. Integrate User Profile

Responsible: TS, SS

Start® End: JUL02 ® JUN03

Dependencies & Risks: 1.2.3; 2.2.1

Goals & Deliverables: Information contained in the user profile will be used to constraint the possible interpretations of the question to those matching the profile.

Description: This task will integrate input from user profile into forming initial question interpretation

Subtask 2.3. Multimedia Dialogue

Responsible: TS, RE

Start® End: JUN03 ® DEC03

Dependencies & Risks: 2.2.1; 4.1, 4.2, 4.3

Goals & Deliverables: Enable mixed media dialogue with English and mouse clicks on the GUI visualization.

Description: Extend Dialogue Manager to handle both language interaction and point & click interaction with visual displays.

Task 3. Information Quality Metrics

Task Responsible: Tang (SUNY), Ng (Rutgers)

Start® End: JAN02 ® DEC03

Dependencies & Risks: 1; 5

Goals & Deliverables: Information Quality Assessment System

Description: This task will develop a system for automated assessment of information quality for various text documents in context of a specific question answering task performed by the user.

Subtask 3.1. Quality Parameters

Responsible: RT

Start® End: JAN02 ® MAR02

Dependencies & Risks: Recruitment of participants in focus group, interview, web survey sessions

Goals & Deliverables: Elicit quality criteria from the previous research and participants. Criteria items are operationalized. Criteria that deem to be crucial to intelligence work will be selected.

Description: Establish the initial set of quality criteria important in intelligence analyst work. Develop a corresponding set of measurable quality indicators that can be computed automatically.

Subtask 3.2. Graphical User Interface tools

Responsible: RE, RT, KBN

Start® End: JAN02 ® JUN02

Dependencies & Risks: 3.1.

Goals & Deliverables: Tools to be developed include: 1) Web survey instrument for quality criteria, 2) Quality assessment interface that allows users to rate quality criteria on TREC documents; 3) Quality assessment Web interface that allows users to rate quality criteria on Web documents

Description: Develop tools for interviewing the users. This includes designing and implementing Graphical User Interface and integrating TREC collection search tools.

Step 3.2.1. Web Survey Instrument for Quality Criteria

Responsible: RT; KBN

Start® End: JAN02 ® MAR02

Dependencies & Risks: none

Goals & Deliverables: A cgi script for collecting quality criteria from Web users.

Description: Design and implement a Web-based script for having Internet users suggest quality criteria they are concerned with when looking for information on the Web.

Step 3.2.2. Quality Assessment GUI for TREC documents

Responsible: RE, RT

Start® End: JAN02 ® MAR02

Dependencies & Risks: none

Goals & Deliverables: A GUI for information quality assessments experiments

Description: Design and implement a graphical user interface to be used by human subjects in information quality assessment experiments.

Step 3.2.3. Quality Assessment GUI for WWW documents

Responsible: RE, RT

Start® End: APR02 ® JUN02

Dependencies & Risks: none

Goals & Deliverables: A GUI for information quality assessments experiments using Web data

Description: Design and implement a graphical user interface to be used by human subjects in information quality assessment experiments.

Subtask 3.3. User experiments

Responsible: RT, KBN

Start® End: JAN02 ® DEC02

Dependencies & Risks: 3.1; 3.2

Goals & Deliverables: Collect quality scores for documents in the test collections

Description: Conduct experiments with users performing specific information tasks. The experiments will be performed in phases with more advanced tools.

Step 3.3.1. Test collection development – Stage 1

Responsible: RT, KBN

Start® End: JAN02 ® MAR02

Dependencies & Risks: none

Goals & Deliverables: Convert TREC queries into questions, select documents with variations of relevance and quality.

Description: Develop test collection from TREC data.

Step 3.3.2. User experiments –Stage 1

Responsible: RT, KBN

Start® End: APR02 ® JUN02

Dependencies & Risks: 3.1, 3.2

Goals & Deliverables: Collect quality scores for TREC documents

Description: Conduct experiments with users performing specific information tasks. The experiments will be performed in phases with increasingly more sophisticated tools.

STEP 3.3.3. Test Collection Development – Stage 2

Responsible: RT, KBN

Start® End: JUL02 ® SEP02

Dependencies & Risks: 3.1, 3.3.1, 3.3.2

Goals & Deliverables: Collection of Web documents

Description: Extend and apply 3.3.2 findings to Web environment.

STEP 3.3.4. User experiments – Stage 2

Responsible: RT, KBN

Start® End: OCT02 ® DEC02

Dependencies & Risks: 3.3.3

Goals & Deliverables: Collect quality scores for Web documents

Description: Conduct experiments with users performing specific information tasks. The experiments will be performed in phases with increasingly more sophisticated tools.

Subtask 3.4. Quality Models

Responsible: TS, RT

Start® End: OCT02 ® DEC03

Dependencies & Risks: 3.1; 3.2; 3.3

Goals & Deliverables: Automated Quality Assessment Module

Description: Use Machine Learning methods to derive models for automated assessment of quality criteria from document features.

STEP 3.4.1. Feature Selection

Responsible: RT; TS

Start® End: OCT02 ® DEC02

Dependencies & Risks: 3.3.1; 3.3.3

Goals & Deliverables: Text features for training quality indicators.

Description: Analyze the quality-annotated documents and identify the text features, such as sentence length, use of vocabulary, use of punctuations, etc. that could serve as quality indicator for an automated system.

STEP 3.4.2. Error-Driven Learning

Responsible: TS; RT

Start® End: JAN03 ® DEC03

Dependencies & Risks: 3.3.1; 3.3.3; 3.4.1

Goals & Deliverables: Train the system to recognize quality through a combination of quality indicators.

Description: Use error-driven learning techniques to induce automated quality assessment capabilities from human-annotated data.

Task 4. Information Visualization

Task Responsible: Erbacher (SUNY)

Start® End: JAN02 ® DEC03

Dependencies & Risks: Sample data

Goals & Deliverables: A visualization and interaction front-end to the question answering and answer fusion environment

Description: Development of novel techniques and capabilities to visually represent clusters and provide for the exploration of said clusters along with the encapsulated documents.

Subtask 4.1. Visual Clustering

Responsible: RE

Start® End: JUL02 ® JUN03

Dependencies & Risks: Sample data

Goals & Deliverables: Visualization techniques representing clusters visually incorporating interaction facilities

Description: Incorporate clustering capabilities in conjunction with the direct manipulation capabilities to probe the clusters for more detailed information and zooming capabilities to view lower level details of the cluster.

Step 4.1.1. Visual Cluster Representation

Responsible: RE

Start® End: JUL02 ® DEC02

Dependencies & Risks: 1.2.3.

Goals & Deliverables: Visualization design and implementation code modules

Description: Provide a visual representation of the cluster groups returned by the data and answer fusion environment

Step 4.1.2. Probing and Selection of Cluster Elements

Responsible: RE

Start® End: JAN03 ® MAR03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Direct manipulation code modules

Description: Perform inverse projection to identify user selected elements within the visualization display. Provide feedback of the selected elements. Allow individual and group selection of the elements for further analysis/processing.

Step 4.1.3. Zooming and Multiple Levels of Representation of Clusters

Responsible: RE

Start® End: APR03 ® JUN03

Dependencies & Risks: 4.1.2

Goals & Deliverables: User interface code modules. Multiple window code modules.

Description: Provide zoomed view of selected clusters as a new window. Provide the ability to view individual documents within a given cluster. Provide additional probing capabilities of individual documents.

Subtask 4.2. Multi-Parametric Data Visualization

Responsible: RE

Start® End: JAN03 ® SEP03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Visualization and interaction techniques to visually represent data parameters as visual attributes.

Description: Develop glyphs to visually represent multi-parametric data, including quality factors, keyword appearance, and data fusion parameters. Investigate ability to incorporate large numbers of attributes in the visual display and the user interface.

Step 4.2.1. Identification of Relevant Data Attributes

Responsible: RE

Start® End: JAN03 ® MAR03

Dependencies & Risks: Sample data and 4.1.1

Goals & Deliverables: Visualization and user interface design

Description: Identify expected data parameters to be mapped to the visual (attributes (glyphs components). Determine the number of parameters to be mapped simultaneously.

Step 4.2.2. Development of Multi-Parametric Visualization Techniques

Responsible: RE

Start® End: APR03 ® SEP03

Dependencies & Risks: 4.2.1

Goals & Deliverables: Visualization code modules

Description: Develop visual techniques to represent the desired number of data parameters as visual attributes. Incorporate the visualization techniques (glyphs) within the visual representation of clusters and document elements.

Step 4.2.3. Parametric Interface Linkage

Responsible: RE

Start® End: APR03 ® SEP03

Dependencies & Risks: 4.2.1

Goals & Deliverables: User interface code modules

Description: Incorporate user interface capabilities to allow user selection of data parameter to visual attribute mappings. Develop an initial set of appropriate mappings and recommendations.

Subtask 4.3. Visual Steering

Responsible: RE

Start® End: APR03 ® DEC03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Ability to visually interact and direct the execution/visualization

Description: Allow the user to dynamically change parameter values during execution. Update the display to reflect the impact in real-time.

Step 4.3.1. Real-Time Interactive Controls

Responsible: RE

Start® End: APR03 ® SEP03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Multi-threaded user interface with continuous updating

Description: Allow user interface parameters and associated data fusion parameters to be changed during runtime. Provide feedback loop by which the changed variable values re-execute the appropriate module of the visualization environment or data fusion module. Controls must respond with real-time requirements. This will require a multi-threaded approach.

Step 4.3.2. Dynamic Updating of the Visual Environment

Responsible: RE

Start® End: APR03 ® SEP03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Re-entrant visualization code

Description: Ensure continuous updating of visual attributes relevant to modified interface controls. Provide selective updating capability and other optimizations. Examine feasibility of real-time continuous modification/updating of the visualization display.

Step 4.3.3. Visually Directed Interactions

Responsible: RE

Start® End: APR03 ® SEP03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Visually based user interface techniques and code modules

Description: Provide the ability to interact directly with the visual display for parameter modification. Parameter to be modified must be selected. Maintain user’s focus on visualization as the display is being updated.

Step 4.3.4. Link Interactive Controls to Visual Environment Parameters

Responsible: RE

Start® End: OCT03 ® DEC03

Dependencies & Risks: 4.3.1, 4.3.2, 4.3.3

Goals & Deliverables: interface code between multiple threads

Description: Directly link parameter elements provided in the user interface to the feedback mechanism and the updating of the visualization display creating a continuous closed loop of execution and interaction.

Subtask 4.4. Multiple Views

Responsible: RE

Start® End: JAN03 ® Dec 03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Multiple views/windows

Description: Provide the ability to maintain multiple representations of the data simultaneously. Incorporate both linked and unlinked views.

Step 4.4.1. Multiple Independent Views

Responsible: RE

Start® End: JAN03 ® JUN03

Dependencies & Risks: 4.1.1

Goals & Deliverables: Multiple view code modules

Description: Incorporate multiple windows with adjustable parameters, individual levels of detail, and individually selected and represented data elements.

Step 4.4.2. View Based Interaction

Responsible: RE

Start® End: JUL03 ® DEC03

Dependencies & Risks: 4.4.1

Goals & Deliverables: Isolated variables and interaction parameters

Description: Provide view (window) based interaction as per sections 4.1.3, 4.2.2, 4.2.3, and 4.3.

Step 4.4.3. Linked Views

Responsible: RE

Start® End: JUL03 ® DEC03

Dependencies & Risks: 4.4.1

Goals & Deliverables: Correlated multiple views

Description: Incorporate the ability to link the interactions between the views and update each view based on the adjustments and interactions performed in a single view providing a correlated exploration of the data environment.

Task 5. Data and Answer Fusion

Task Responsible: Kantor, Strzalkowski (Rutgers, SUNY)

Start® End: JAN02 ® DEC03

Dependencies & Risks:

Goals & Deliverables: Evidence fusion system; Answer fusion module

Description: Develop capabilities for evidence fusion from multiple sources. This is required to increase confidence in the results of initial information search. Fusion is based on relevance and quality of sources. Furthermore, develop techniques of generating answers fused from fragments obtained from multiple documents.

Subtask 5.1. Fusion parameters

Responsible: PK, KBN

Start® End: JAN02 ® DEC02

Dependencies & Risks: 3

Goals & Deliverables: Various fusion functions for different purposes and emphases.

Description: Develop some alternative parameter sets for data fusion (based on precision, and on scheme difference; allowing for user tuning of weights).

Step 5.1.1. Multiple Search Engines

Responsible: PK, KBN

Start® End: JAN02 ® MAR02

Dependencies & Risks: none

Goals & Deliverables: Several IR engines for fusion experiments

Description: This task is to set up several search engines running in order to enable fusion experiments. All engines working on TREC collection.

Subtask 5.2. Automatic Adjustment of Fusion parameters

Responsible: PK

Start® End:

Dependencies & Risks: 3, 5.1 No risks

Goals & Deliverables:

Description: Based on the experiments with real subjects, and feedback through the interface, develop methods to adjust fusion parameters to adapt to a specific user, as searches progress, through combining the results of multiple retrieval schemes.

Step 5.2.1. MAchine learning of Fusion parameters

Responsible: PK

Start® End: MAR02 ® MAR03

Dependencies & Risks: 4.1, 4.2, 4.3, 4.4, 5.1

Goals & Deliverables: Assessed and prioritized algorithms for adapting to user preferences through data gathered in the GUI.

Description: Machine learning algorithms will be applied to estimate what values of the tunable parameters most satisfy the needs of a specific user. This will be done by background optimization as the user's assessment of specific documents and answer sets is learned iteratively. The best fusion is the one that produces the tightest concentration of relevant documents or answer sets in the users preferred display layout.

Subtask 5.3. Answer Fusion

Responsible: TS, TL

Start® End: MAR02 ® JUN03

Dependencies & Risks: no dependencies. A critical component for generative answers.

Goals & Deliverables: The goal is to develop a text fusion method for composing well-formed answers from multiple text fragments of varying content and quality.

Description: Develop an approach and method for automated answer fusion from multiple sources of relevant information adapting the experimental technique for text reduction and summary fusion.

Step 5.3.1. Text Alignment By Content Reduction

Responsible: TL, TS

Start® End: MAR02 ® SEP02

Dependencies & Risks: none

Goals & Deliverables: Align a summary report with the source documents to show the most likely way of how the summary was derived.

Description: This task will develop a Hidden Markov Model that will estimate the maximum probability that a given sequence of words were chosen to produce a cumulative report from several sources. The method will explain when any extra-textual reductions and expansions exist in the summary. We will use a set of training data from New York Times and from the DUC evaluation.

Step 5.3.2. Generative Text Fusion

Responsible: TL, TS

Start® End: SEP02 ® JUN03

Dependencies & Risks: 5.3.1

Goals & Deliverables: The goal is to produce a summary/response by fusing several source texts.

Description: Train the model build in Step 5.3.1 on a large collection of sources and answers and then apply to new texts. Use human judges to correct mistakes and retrain using error-drive learning.

Task 6. Evaluation and User Studies

Task Responsible: Kantor, Ng, Tang (Rutgers, SUNY)

Start® End: JAN02 ® DEC03

Dependencies & Risks: 3,5.1,5.2

Goals & Deliverables: Assessment of alternative schemes for retrieval, for fusion, and for adaptation to the users behavior and preferences.

Description: Perform efficient designed experiments to identify most promising schemes for retrieval, fusion and adaptation. Provide an overall final assessment of all the systems studied.

Subtask 6.1. Formative Evaluation

Responsible: PK, KBN

Start® End: JAN03 ® MAR03

Dependencies & Risks: 3; 4; 5

Goals & Deliverables: Initial combinations of design and user-manipulative variables.

Description: Perform rapid (2-month) cycles of prototyping and testing to explore the space of design options and move in the most promising directions.

Step 6.1.1. Usability Testing

Responsible: RT, KBN

Start® End: JAN03 ® MAR03

Dependencies & Risks: 3,4,5

Goals & Deliverables: The goal is to assess specific features and their influence on effectiveness. Reports of system performance and usability; design recommendations.

Description: Conduct usability experiments to evaluate the system performance and interface design

Subtask 6.2. Summative Evaluation

Responsible: PK, KBN

Start® End: JAN03 ® DEC03

Dependencies & Risks: 3,4,5

Goals & Deliverables: The goal is to assess several possible designs for the system. The deliverable will be a report including a ranking or other relative assessment of the alternatives, in terms of both functionality and usability.

Description: In year 2 we will move also to summative evaluation, to establish the effectiveness of the system developed, as a tool to assist the work of intelligence analysts.

Task 7. Project Management

Task Responsible: Strzalkowski (SUNY)

Start® End: JAN02 ® DEC03

Dependencies & Risks: none

Goals & Deliverables: Reports, presentations, demonstrations

Description: Periodic reporting to the Sponsor, project reviews, project meetings.

Subtask 7.1. Progress Reporting

Responsible: TS and all

Start® End: JAN02 ® DEC03

Dependencies & Risks: none

Goals & Deliverables: periodic performance and progress reports on monthly and quarterly basis.

Description: Use sponsor-provided templates to report monthly progress and financial data

Subtask 7.2. Project Reviews

Responsible: TS and all

Start® End: JAN02 ® DEC03

Dependencies & Risks: none

Goals & Deliverables: periodic performance reviews on quarterly basis

Description: On –site reviews will be performed on the request of the sponsoring agency at approximately 3 month intervals.

Subtask 7.3. Project Meetings

Responsible: TS and all

Start® End: JAN02 ® DEC03

Dependencies & Risks: none

Goals & Deliverables: Project meetings organized by ARDA

Description: Attend all program meetings as required by ARDA at approximately 6-month intervals. At least the PI and the subcontract PI must attend all meetings.

 

 

 

Schedule

Schedule of Milestones and Deliverables

Timing

Milestones and Deliverables

TASK

monthly

  • Monthly progress reports
  • 7.1

MAR02

  • Quality criteria selected
    • CGI-Script for collecting quality criteria
    • GUI for IQ assessment experiments
    • TREC questions selected
    • Multiple retrieval engines working
  • Quarterly Report #1

  • 3.1
  • 3.2.1
  • 3.2.2
  • 3.3.1
  • 5.1.1
  • 7.2

JUN02

    • Question Processing Module
    • Attribute structure for questions
    • Text chunker

  • All GUI tools developed
    • GUI for IQ experiments of Web data
    • Quality scores experiments completed

  • Quarterly Report #2
  • 6-month program meeting
  • 1.1.1
  • 1.2.1
  • 1.2.2
  • 3.2
  • 3.2.2
  • 3.3.1
  • 7.2
  • 7.3

SEP02

    • Topical clustering module

  • Semantic Interpretation Module

    • Quality scores for Web documents
    • Source to answer alignment

  • Quarterly Report #3

  • 1.2.3
  • 1.2
  • 3.3.3
  • 5.3.1
  • 7.2

DEC02

    • Question typology adopted

  • Question Processing and Typing Module
    • Cut-and-paste answers
    • Quality assessments in answers
    • Web documents collected
    • Quality scores for Web documents

  • User experiments completed
    • Quality indicators selected
    • Visualization of Clusters

  • Evidence fusion functions established
  • Year 1 Report
  • Year 1 Review & meeting

  • 1.1.2
  • 1.1
  • 1.4.1
  • 1.4.2
  • 2.1.1
  • 3.3.4
  • 3.3
  • 3.4.1
  • 4.1.1
  • 5.1
  • 7.2
  • 7.3

MAR03

    • Probing and selection of clusters
    • Identification of Data Attributes
    • Usability evaluation completed

  • Formative Evaluation completed
  • Quarterly Report #4

  • 4.1.2
  • 4.2.1
  • 6.1.1
  • 6.1
  • 7.2

JUN03

    • Text chunk filter
    • Cluster signatures
    • Theme Router
    • User profiles

  • Dialogue Manager
    • Generative text fusion

  • Visual Clustering Completed
    • Cluster Zooming with Multiple Levels
    • Multiple Independent Views

  • Answer Fusion module
  • Quarterly Report #5
  • 18-month program meeting

  • 2.1.1
  • 2.1.2
  • 2.2.1
  • 2.2.2
  • 2.2
  • 5.3.2
  • 4.1
  • 4.1.3
  • 4.4.1
  • 5.3
  • 7.2
  • 7.3

SEP03

  • Multi-Parametric Data Visualization Completed
    • Multi-Parametric Visualization
    • Parametric Interface Linkage
    • Real-Time Interactive Controls
    • Dynamic Environment Updates
    • Visually Directed User Interactions

  • Quarterly Report #6

  • 4.2
  • 4.2.2
  • 4.2.3
  • 4.3.1
  • 4.3.2
  • 4.3.3
  • 7.2

DEC03

  • HITIQA Clarification Dialogue
  • HITIQA Answer Generator

HITIQA QA System

  • Multimedia Dialogue

Dialogue Management System

    • Automated quality assessment

  • Information Quality Models

Information Quality Assessment System

Information visualization

  • Visual Steering Completed
    • Link Interactive Controls to Display

  • Multiple Views Completed
  • Summative evaluation completed
    • View Based Interaction
    • Linked Views

Evaluations and user studies completed

  • Final Report
  • 24 month Meeting & Review

  • 1.3
  • 1.4
  • 1
  • 2.3
  • 2
  • 3.4.2
  • 3.4
  • 3
  • 4
  • 4.3
  • 4.3.4
  • 4.4
  • 4.4.2
  • 4.4.3
  • 6.2
  • 6
  • 7.2
  • 7.3