CAREER: Intelligent Generation of Text and Information Graphics - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

CAREER: Intelligent Generation of Text and Information Graphics

Description:

General explanation of background on genetics. Counseling role ... complexity of causal explanation. emotionally disturbing information ... – PowerPoint PPT presentation

Number of Views:37
Avg rating:3.0/5.0

less

Transcript and Presenter's Notes

Title: CAREER: Intelligent Generation of Text and Information Graphics


1
CAREER Intelligent Generation of Text and
Information Graphics
  • Motivation
  • vital technical information involving scientific
    or medical arguments may be difficult for lay
    person to grasp
  • Proposal
  • use AI to help technical experts produce
    user-friendly arguments in text and/or graphics
  • use HCI methods to ensure effectiveness
  • build demonstration system (GenIE) for genetic
    counselors

This material is based upon work supported by
the National Science Foundation under Grant No.
0132821. Any opinions, findings, and conclusions
or recommendations expressed in this material are
those of the author and do not necessarily
reflect the views of the National Science
Foundation.
2
The Genetic Counselor
  • Meets with clients
  • Informational/educational role
  • Explanation of diagnosis of genetic condition
  • Explanation of inheritance risks
  • General explanation of background on genetics
  • Counseling role
  • Writes summary letter (1-2 pages) for client

3
Client Issues
  • Complex Subject
  • probability and statistics
  • hypothetical outcomes
  • causality
  • scientific and medical terminology
  • diagrams may help
  • Emotional Distress
  • Readers ability to comprehend (innumeracy)
  • Rapidly changing information

4
GenIE Genetics Intelligent Editor
  • Goals
  • NOT to replace human counselor
  • Reduce counselors effort
  • artificial intelligence creates 1st draft of
    letter
  • human counselor may revise or reject GenIEs
    draft
  • Design online presentation client benefits
  • supplementary graphics and animation
  • links to other resources
  • automatic updates

5
Multiple Research Methods
Goal Evaluate presentation techniques under
controlled conditions in lab
HCI Experiments
Computational Model Building
GenIE
Goal concrete implementation of ideas for
demonstration and evaluation
Goal develop widely applicable computational
(AI) techniques for generating arguments
Corpus Analysis
Goal study corpus to understand how human
authors communicate technical arguments
6
Research Methods HCI Experiments
  • Before Formally evaluate effectiveness of
    communication techniques before computational
    models created, e.g.
  • How does layout of document affect comprehension
    of arguments?
  • What types of information to present in text, in
    graphics, or both?
  • Graphical depiction of argument structure
  • After Evaluate communicative effectiveness of
    presentations created by GenIE
  • ablation experiments to identify which factors
    contribute or detract from communicative
    effectiveness

7
Research Methods Corpus Analysis
  • Corpus Acquisition (text and graphics)
  • genetic counseling summary letters, client
    education documents (print and web)
  • Qualitative Analysis
  • types of information graphic techniques
  • analysis of argumentation (ex. predictive,
    diagnostic, value-based, Toulmin-style,
    dialectical)
  • Computational Linguistics Analysis
  • develop coding scheme with intercoder reliability
  • manually encode corpus
  • manual and automated discovery of communication
    techniques evolved by human authors

8
Research Methods Computational Models
  • Develop AI methods to
  • represent the underlying scientific arguments and
    reasoning of the experts
  • predict the audiences potential problems in
    understanding, e.g.,
  • complexity of causal explanation
  • emotionally disturbing information
  • reason about content (both text graphics),
    organization, and layout to avoid predicted
    problems
  • generate text and graphics based on above

9
Analysis of Argumentation in Corpus
  • Argumentation discourse that weighs evidence and
    presents multiple points of view
  • An important dimension of argumentation in
    letters in corpus diagnostic and predictive
    reasoning
  • hearing loss was caused by mutation in gene
    (GJB2)
  • if HD, then chance that others in family are
    affected
  • Those parts of letter can be represented by
    Bayesian (belief) network

10
Bayesian Network
History/proband Age child
History/mother family history of deafness no
Genotype/mother one abnormal copy of gene GJB2
Genotype/father 2 abnormal copies of gene GJB2
50
50
Genotype/proband 2 abnormal copies of gene GJB2
Genotype/sibling 2 abnormal copies of gene
GJB2
Biochemistry/proband Connexin 26 abnormal
Physiology/probandnormal chemical equilibriumno
Symptom/proband deafness
Finding/proband facial defects no
Result/proband GJB2 test positive
Symptom/father deafness
Symptom/sibling deafness
11
GenIE Project Summary
  • Building demonstration system (GenIE) to help
    genetic counselors write letters
  • Using HCI to ensure effectiveness of general
    argument presentation techniques
  • Using AI to to model experts reasoning and
    argumentation strategies
  • Techniques will be applicable in many domains to
    problem of computer-assisted or automatic
    multimedia generation of effective technical
    arguments for lay audience
Write a Comment
User Comments (0)
About PowerShow.com