hitiqa2_vol1 - School of Communication and Information

advertisement
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
VOLUME 1 - Technical / Management Details
(40 pages max excluding covers)
Organization / Company
CAGE Code
DUNS / CEC Number
TIN Number
Type of Business
Proposal Title
University at Albany, SUNY
OTHER EDUCATIONAL
System Design
Perspective Category
(Check ONLY ONE Box)
(See Section 5.1)
_X_ 1. End-to-End System
___ 2. Component Elements
___ 3. Cross Cutting / Enabling Technologies
If "Component Elements"
Category Selected Above
(Check ALL that Apply
(See Section 5.1.2)
___
___
___
___
Data Strategy
(See Section 5.2.3)
___ Focused Data Strategy
_X_ Diverse Data Strategy
___ Other (Identify):
HITIQA-2
Question Understanding and Interpretation
Determining the Answer
Formulating and Presenting the Answer
Other (Identify):
Page 1
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
BAA 03-06-FH - VOLUME 1 - Technical / Management Details (CONTINUED)
Team Members / Type of
Business
University at Albany, State University of New York /
Other Educational
Rutgers, State University of New Jersey /Other
Educational
City University of New York, Lehman College /other
Educational
Principal Investigator(s)
Name(s)
Mail Address
Professor Tomek Strzalkowski
Phone Number
Fax Number
E-mail Address
Administrative Contact
Name
Mail Address
518-442-2608
518-442-2606
tomek@csc.albany.edu
Ms. Linda Donovan
Phone Number
Fax Number
E-mail Address
518-437-4555
Proposal Duration
Cost - Year 1
Cost - Year 2
Total Cost
24 months
$
$
$
HITIQA-2
University at Albany, SUNY
1400 Washington Avenue, SS-262
Albany, NY 1222
ldonovan@uamail.albany.edu
Page 2
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
Part I: Summary of Proposal. (Tomek + Paul)
National security depends more than ever upon accurate, high-quality information
being available at the right time to support important policy decisions. A key element of
this process is the work of the intelligence analyst, who must quickly and efficiently
produce the right information from a potentially enormous number of sources, reports,
documents and databases. Today's information retrieval and factoid question answering
technology provides some help, but is also a source of growing frustration and missed
opportunities. There is little doubt that a more powerful technology is needed to address
question answering needs of professional intelligence analysts. The first phase of
ARDA’s AQUAINT Program has pushed the QA technology out of its infancy into the
realm where we can start seeing its benefits in the future.
The HITIQA team has made a significant progress in Phase 1 of AQUAINT. We have
developed a prototype analytical QA system in which we demonstrated preliminary
solutions to several key goals of the AQUAINT program:
 Accepting complex, analytical questions in a form natural to the analyst.
 “Understanding” the questions in context of available unstructured data.
 Negotiating this understanding with the analysts through a multimodal dialogue.
 Providing access to related adjunct information uncovered through the search and
framing process.
 Delivering to the analyst means for exploring the answer space via interactive
visualization.
 Generating a preliminary answer out of fused “headlines” and fragments of
original source material
In addition, we have completed extensive studies in several aspects of usability and
efficacy of the emerging QA technology and its use in the intelligence environment:
 A series of usability studies, including hands-on sessions with the USNR and
USAF analysts, confirmed the strengths of overall HITIQA approach, while also
exposed limitations of the current prototype. Web-based implementation made the
technology available to a variety of users who provided feedback and helped us to
capture their expectations.
 Two series of information quality experiments have produced a large volume of
data collected from dozens of users over many sessions and across different
backgrounds, leading to a preliminary conclusion that the perception of
information quality is an acquired and highly individualized phenomenon.
 Similarly, experiments with information fusion revealed that
HITIQA technology thus represents a radical departure from the “factoid” question
answering that has dominated the research landscape until now. While factoid systems
have made significant strides in accuracy and efficiency, as demonstrated by the most
recent TREC QA evaluations, their utility has always been limited in the world of
professional analysts. Therefore, based on our successes with HITIQA-1 system and our
growing experience and understanding of the analytical process, and given the goal of
AQUAINT Phase 2, we propose another radical leap forward to turn HITIQA from a
helper tool into an indispensable, highly adaptable and personalizable “analyst’s
HITIQA-2
Page 3
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
assistant”. In doing so, we will address the following Phase 2 goals, schematically
illustrated in Figure 1:







Question Answering as Part of a Larger Information Gathering Process: HITIQA2 will support the analyst throughout the entire analytical process associated with
an “analytical scenario”. This means not only accepting a series of interrelated
questions but also providing their interpretation in context of the overall
information task.
Accessing, Retrieving and Integrating Diverse Data Sources: HITIQA-2 will
exploit structured data as a source of pre-processed information of direct interest
to the analyst, as well as a source of knowledge that can adapted to provide a
better understanding of unstructured, unprocessed, novel information.
Interact with the system, using questioning strategies natural to the analyst: We
will advance current triaging dialogue and visual browsing in HITIQA-1 to full
problem-solving dialogue and exploratory navigation that will provide
cooperative environment where the system actively assists the analyst in his/her
work.
Explore boundaries of statistical and linguistic approaches to QA: HITIQA is
already a hybrid system encompassing a variety of statistical and linguistic
methods for information processing. This will be significantly expanded by
adding knowledge acquisition methods that will utilize structured databases to
learn how to process unstructured data with accuracy comparable to manually
built knowledge-based methods, while also scalable to new and diverse domains.
Adapt to analyst’s preferred problem solving style: We will build into HITIQA
automated mechanism for adapting the system’s performance to closely match the
analyst’s personal preferences and style. This will be achieved over time through
an adaptation process that tracks analyst’s information selections and interaction
patterns and adjusts system’s behavior accordingly.
Maintain analyst’s confidence in the QA process: HITIQA-2 will create and
maintain a persistent network of successive models reflecting the analyst’s
information exploration strategy and a changing peripheral context. This will
include a working space of the currently active answer model, as well as the
backdrop of secondary information which can be explored to guarantee
completeness.
Evaluating, Validating and Presenting the Answer: In HITIQA-2, the answer, in
the form of a preliminary analytical report, will be assembled from the structured
knowledge sources and unstructured data items. This will be accomplished
through adoption of frame-based semantics, shared among multiple data sources.
HITIQA-2
Page 4
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
Figure 1. HITIQA-2 Concept and Components
A.
Summary of Innovative Claims
HITIQA has been conceived as a long term research project to address the challenges for
the intelligence community identified in the AQUAINT Program as a whole. In Phase 1
we attacked a number of these challenges, finding solutions to some and making inroads
into others. We have also discovered additional challenges that need to be solved before
the QA technology can have visible impact on the work of the intelligence analyst. What
we propose for Phase 2 is therefore not an incremental addition to our Phase 1 work;
rather the challenges before us require an entire new set of innovations to be delivered.
These innovations are summarized in Table 1 below by laying out Phase 1 advances and
proposed Phase 2 goals against the grid of overall objectives for HITIQA project.
Table 1: HITIQA-2 Advances compared to HITIQA-1 base
HITIQA
INNOVATIONS
PHASE 1
PHASE 2
Questions
 Single analytical
 Scenarios involving series of
questions and strategy
Dialogue
 Clarification triage
 Clarification
 Navigation and Problem-
HITIQA-2
Page 5
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
solving
Answers
 Fused headlines and text
passages
 Fused passages and generated
reports
Semantics
 Data-driven in general
domain
 Manual fit over specialized
domain
 Data and knowledge-driven,
 Domain adaptable
 Knowledge acquisition from
structured sources
Task-level
persistence &
adaptability
 Not adaptable
 One model per interaction
 Feedback with source fusion
 Successive models following
analyst’s task strategy
 Model backdrop context
User-level
persistence &
adaptability
 None
 No personalized features
 Adapts to user information
selection and judgments
 Keeps memory of interactions
 Always on if needed
Visualization
 Answer space topology
 Interaction alternative to
dialogue
 Event and relationship map
 Navigation/exploration
 Coordinated multimodal
interaction
Information
Quality &
Usability
 Measured per source
 9 empirical quality criteria
 Measured per source & topic
 Individualized criteria based on
the analyst’s pattern of use
Evaluations &
Usability
Studies
 Program-level pilots
 Short-sessions with users
 USNR sessions
 Program-level metric-based
evaluations
 Sustained usability testing with
USNR, USAF, other analysts
Data sources
 Unstructured text
 Unstructured text
 Structured databases
 Web based sources
B.
Summary of Technical Rationale
Brief summary of the technical rationale, technical approach, and constructive plans for
accomplishment of technical goals.
The key technical challenge to developing a practical QA system for the intelligence
analyst is equip it with capacity to substantially assist the analytical process. This means
being able to augment analyst’s capabilities for locating and correlating information,
detecting information of a certain kind etc, and doing so without prior detailed knowledge
of the topic. Achieving this requires an imaginative combination of existing knowledge
and projecting it over the new data.
--- structured data can be converted into knowledge
HITIQA-2
Page 6
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
--- this knowledge can be projected over novel information to achieve an initial, partial
structuring
--- this structuring is can be used to derive extraction tools to locate further items of
interest in unstructured data
--- the partial understanding of the data can be refined through the dialogue with the
analyst so that (1) a model of answer space is created that can be navigated, (b) the
system grasp of the task domain semantics is refined, and (c) a revised model is built.
--- the system’s growing grasp of analyst’s goal and strategy is projected from the refined
model into the larger data context in the form of data fusion (based on analyst’s perceived
usefulness of information rather than any specific “objective” metric such as relevance).
--- all these observed data are rendered into 3-D interactive visualization that provides an
orthogonal interaction mode to the language based dialogue
•
•
•
•
Scenario-based QA Interaction
– Complete problem solving support
– Builds successively more accurate answer space models
– Detects when full answer space seen
Adaptable Analyst’s Personal Assistant
– Adapts to analyst’s background and style
– Keeps memory of interactions, tasks and solutions
– Optimizes source selection for relevance & quality
Instant Domain “Expertise” and Adaptation
– Adapt and absorb structured data sources as knowledge
– Optimize through bootstrap learning
– Project knowledge structure over unstructured sources
Tryouts and Evaluation
– Metrics evaluations on answer completeness and compactness
– Focus on usability testing (e.g. USNR and USAF analysts)
– Integrated multi-modal interaction and navigation
Box 1 HITIQA-2 Key Research Objectives
C.
Schedule and Milestones
Schedule and milestones for the proposed research, including overall estimates of cost for
each task. A one-page graphic illustration that depicts major milestones of the proposed
effort arrayed against the proposed time and cost estimates must be included.
D.
Summary of Deliverables
A summary of the deliverables associated with the proposed research.
E.
Key Personnel
A clearly defined organizational chart of all anticipated program participants with brief
biographical sketches of key personnel and significant contributors, their roles (including
HITIQA-2
Page 7
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
role of Principal Investigator) and their level of effort in each year (calendar year or
academic / summer year) of the program. A chart, such as the following, is suggested.
Participants
Organization
Role
Year 1
Year 2
Prof. Tomek Strzalkowski
University at Albany
Key Personnel/
PI, PM
25%
25%
Prof. Deborah Andersen
University at Albany
Significant
Contributor
25%
25%
Ms. Sharon Small
University at Albany
Significant
Contributor
100%
100%
Doctoral Candidate 1
University at Albany
Contributor
50%
50%
Doctoral Candidate 2
University at Albany
Contributor
50%
50%
Graduate Assistant 1
University at Albany
Contributor
50%
50%
Graduate Assistant 2
University at Albany
Contributor
50%
50%
Graduate Assistant 3
University at Albany
Contributor
50%
50%
Prof. Paul Kantor
Rutgers University
Key Personnel/
co-PI
25%
25%
Prof. Nina Wacholder
Rutgers University
Significant
Contributor
25%
25%
Prof. K.B. Ng
Rutgers University
Significant
Contributor
25%
25%
Graduate Assistant 1
Rutgers University
Contributor
50%
50%
Graduate Assistant 2
Rutgers University
Contributor
50%
50%
Prof. Boris Yamrom
City University of
New York
Key Personnel
25%
25%
Graduate Assistant 1
City University of
new York
Contributor
50%
50%
Professor Tomek Strzalkowski – University at Albany, SUNY
Education: Simon Fraser University, PhD Computer Science, 1986.
Experience: Dr. Strzalkowski is an Associate Professor of Computer Science at SUNY
Albany. Prior to joining SUNY, he was a Natural Language Group Leader and a Principal
Computer Scientist at GE CRD. Prior to GE, he was an Assistant Professor of Computer
Science at New York University. He received his PhD in Computer Science from Simon
Fraser University in 1986 for work on the formal semantics of discourse. He has done
research in a wide variety of areas in computational linguistics, including database query
systems, formal semantics, and reversible grammars. He has directed research projects in
natural language processing and information retrieval sponsored by ARDA, DARPA and
NSF, including work under several TIPSTER contracts. While at GE, he was developing
advanced text summarization systems for the Government. Dr. Strzalkowski has
published over a hundred scientific papers on computational linguistics and information
HITIQA-2
Page 8
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
retrieval. He is the editor of two books: Reversible Grammar in Natural Language
Processing, and Natural Language Information Retrieval. Current sources of support
include DARPA-funded AMITIES project (2001-04; 20% commitment) and ARDAfunded cross-document summarization project (2000-02; 20%). Pending proposals:
DARPA EELD (2001-06; 20%) and NSF ITR (2001-04; 10%).
Professor Paul B. Kantor (http://scils.rutgers.edu/~kantor) Rutgers University
Education: Ph.D. Theoretical Physics, Princeton University (1963)
Experience: Dr. Kantor is Professor of Information Systems in the School of
Communication, Information and Library Studies at Rutgers, the State University of New
Jersey. Previously he served as a faculty member at Case-Western Reserve University, in
the departments of Physics, Library Science, System Engineering, and Operations
Research. At Rutgers since 1991, he has directed numerous research projects on the
development and evaluation of library and information systems, most notably the ANLI
system for augmenting a library online catalog with hyperlinks, and the AntWorld
project. Prof. Kantor is also a Member of the internationally renowned Rutgers Center for
Operations Research (RUTCOR), director of the Alexandria Project Laboratory, and
director of the Rutgers Distributed Laboratory for Digital Libraries. He is author of more
than 160 journal articles, book chapters, conference papers and technical reports, and his
research has been supported by the ONR, the Institute for Defense Analysis, NSF,
DARPA, and other organizations. He is a regular participant in the NSF Information and
Data Management planning conferences, serves as a reviewer for numerous scientific and
scholarly journals, and is a Fellow of the American Association for the Advancement of
Science and the founding Editor in Chief of the journal Information Retrieval. Current
projects include the DARPA-funded Novel Approach to Information Finding (AntWorld)
N66001-97-C-8537 (15%). Pending projects include Dynamic Indexing and Archiving
of Brain Images (NSF/ITR 11%) and Disruption of Quantum Coded Messages (NSF
ITR/SY. 15%)
<ADD ALL key personnel and significant contributors>
HITIQA-2
Page 9
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
Part II: Detailed Proposal Information.
This part shall provide the detailed, in-depth discussion of the proposed research.
Specific attention must be given to addressing both the risks and payoffs of the proposed
research making it desirable to pursue. This Part shall provide:
A.
Innovative Claims (Tomek, Paul, Boris)
This is the centerpiece of the proposal and should succinctly describe the unique
proposed contribution.
 Question Answering as Dialogue with Heterogeneous Data
 Instant Domain Expertise: Knowledge acquisition from structured data
 Knowledge bootstrapping and induction
 Scenario based extended interaction support – creation of multiple models, model
evolution, model revision – sessions-based persistence
 Personalization of interaction style and personal preferences – user-based
persistence. Learning from choices made: optimizing all aspects of the system –
retrieval, fusion, quality, framing, answer.
 Self adaptation: create good-enough initial system, improves with usage
 Answer generation: from headlines to reports
 Data integration across sources – frame based
 Visualization: exploring navigation dimension – 3D, event-relationship map,
time; tightly coupled with dialogue
 Information quality – aspects? Attitudes? – Quality Model derived for an analyst
– multiple models can be used to provide various viewpoints
 Information fusion: view outside of the model window, zoom in on out parts
means more retrieval and new model.
B.
Detailed Technical Rationale
The technical rationale should clearly show why the proposed technical approach is
expected to achieve the stated purpose within the proposed cost and time schedule. The
rationale shall also describe the rationale for the claims and deliverable products outlined
elsewhere in the proposal and show how past / current performance justifies an award in
this technical area.
Background on HITIQA-1 advances and accomplishments, data-driven semantics of
questions, clarification dialogue (triage) etc. All this is assumed “solved” and available as
building blocks for Phase 2.
B.1. Question Answering as Multimodal Dialogue with Heterogeneous
data Sources (Tomek)
Adaptation of structured data into knowledge
Interactive QA from structured data
Projecting structures over unstructured data
HITIQA-2
Page 10
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
B.2. Scenario level interaction using evolving successive models
(Tomek)
Follow up questions, continuations, changes of direction, detours, drill-downs
Model structure – frames and links
B.3. Acquiring domain knowledge from structured data (Tomek)
Instant frames from structured data
Validation against current data and task
Bootstrapping and tuning from the task data
B.4. Self-adaptation and Personalization (Tomek, Paul, Nina)
Adaptation to task and approach – capturing a strategy
Adaptation to the analyst interaction style
Selecting and loading other styles and strategies
Adaptable source fusion
B.5. Advanced Answer Generation (Tomek)
Generation of reports from frames, models, and structured data
B.6. Information Quality Models (Paul, Nina, KB)
Personalized quality models
Quality model selection and exchange
B.7. Navigating Answer Space through 3-D interactive visualization
(Boris)
Model navigation, model background, moving lens
Controlling background processes
C.
Statement of Work
SOW describing the effort’s scope, the specific tasks to be performed and their associated
schedules. At a minimum, the statement of work shall consist of the following sections:
C.1. Scope of the Proposed Work
A statement as to what the SOW covers: objectives and goals and major milestones for
the effort. Key elements are task development and deliverables.
TABLE 2. SCHEDULE OF MILESTONES AND DELIVERABLES
TIMING
monthly
3 month
6 month
HITIQA-2
MILESTONES


Page 11
DELIVERABLES
Status report
Quarterly Report #1
Quarterly Report #2
3/9/2016
BAA 03-06-FH AQUAINT Phase II
SUNY Albany/Rutgers/CUNY
8 month

9 month

10 month

12 months

15 months

HITIQA version 1.0
Year 1 Report & Review
Quarterly Report #4
18 months

Quarterly Report #5
21 months

Quarterly Report #6
24 months
 Final system assembly
 Components API’s
HITIQA version 2.0
Final Report & Review
Quarterly Report #3
C.2. Technical and Task Requirements
A description of tasks, representing the work to be performed, developed in an orderly
progression and in enough detail to establish the feasibility of accomplishing the overall
program goals. The overall effort should be grouped into major tasks and identified in a
work breakdown structure (WBS)-like numbering system. Proposed costs shall have a
one-to-one correlation to this reporting structure, which shall be depicted in the cost
volume.
Task 1
Task 2
D.
Description of Results
A description of the results, products, transferable technology and an expected
technology transfer path must be included.
E.
Comparison with On-going Research,
Highlighting the uniqueness of the proposed effort / approach and differences between
the proposed effort and current state-of-the-art clearly stated. Identify the advantages and
disadvantages of the proposed work with respect to potential alternative approaches.
Compare with other AQUAINT work (LCC, IBM, NMSU, ISI…)
Compare with other QA works
HITIQA-2
Page 12
3/9/2016
BAA 03-06-FH AQUAINT Phase II
F.
SUNY Albany/Rutgers/CUNY
Previous Accomplishments
Discussion of Offeror's previous accomplishments / work in this or closely related
research areas.
G.
Description of Facilities
that would be used for the proposed effort.
H.
Statement about Government-furnished Property
If any portion of the research is based on the use of Government-owned resources of any
type, the Offeror shall specifically identify the property or other resource required, the
date the property or resource is required, the duration of the requirement, the source from
which the resource will be obtained, if known, and the impact on the research if the
resource cannot be provided. If no Government-furnished property is required for
conduct of the proposed research, this section shall consist of a statement to that effect.
I.
Proposed Research Team
Detailed description of the support, including formal teaming agreements, required to
execute the Offeror's proposal. Discussion of teaming relationships should include the
programmatic relationship of team members; the unique capabilities and relevant
accomplishments and concise summary of qualifications of all team members (key
personnel and significant contributors), with information about their major sources of
support and commitments of their time; the task responsibilities of team members; the
teaming strategy among the team members; and the management approach for the team.
Full resumes / curriculum vitae of key personnel and significant contributors should be
included in Volume 2 (Additional Reference Information) of the proposal.
J.
Proprietary Claims
A summary of any proprietary claims to results, prototypes, or systems The Offeror
shall submit a separate list of all technical data or computer software that will be
furnished to the Government with other than unlimited rights in accordance with DFARS
252.227-7017, Identification and Assertion of Use, Release or Disclosure Restrictions.
All AQUAINT contractors will be required to provide deliverables (software and
documentation) for integration with other AQUAINT Program contractor’s products for
use in testbed evaluations and demonstrations in an end-to-end simulated operational
environment. (See Section 5.4 for more information about the testbed.)
K.
Proposed Evaluations
Description of how progress toward completion of their research goals will be measured,
including a description of the evaluations to be performed, a schedule of implementation
and type of report to be prepared.
HITIQA-2
Page 13
3/9/2016
BAA 03-06-FH AQUAINT Phase II
L.
SUNY Albany/Rutgers/CUNY
Data Sources
Identification and description of anticipated data sources to be utilized in pursuit of the
project research goals.
AQUAINT, CNS, WMD
Additional structured databases from CNS, USGS, …
M.
Participation in AQUAINT Testbed
Summary of a plan, schedule and process for participation in the AQUAINT testbed.
HITIQA-1 system is being currently deployed at the MITRE testbed. We plan that the
HITIQA-2 prototype could be inserted at the end of first year of Phase 2. We assume that
HITIQA-2 will be backward compatible with HITIQA-1 thus allowing insertion of
selected components as soon as they are developed.
HITIQA-2
Page 14
3/9/2016
Download