from question to query: an intelligent strategy for making complex data accessible to novice users...

Click here to load reader

Post on 18-Dec-2015




0 download

Embed Size (px)


  • Slide 1
  • From Question to Query: An Intelligent Strategy for Making Complex Data Accessible to Novice Users Albert Anderson, Public Data Queries, Inc. Edward Brent, Idea Works, Inc. Pawel Slusarz, Idea Works, Inc. Heather Branton, Public Data Queries, Inc;. IASSIST 2003 Ottawa Strength in Numbers May 30, 2003
  • Slide 2
  • Acknowledgements PDQ-Expert is based on an integration of: Qualrus, a qualitative analysis system from The Idea Works; PDQ-Explore, an information retrieval system from Public Data Queries, with contributions from: Paul AndersonJohn Vidolich Patrick RadyMarc Williams Lisa Neidert, Consultant
  • Slide 3
  • And Thank You to NSF and NIH The Idea Works has developed the Qualrus system for qualitative analysis in part with the support of small business funding from the National Science Foundation (NSF). Public Data Queries, Inc., has developed PDQ-Explore in part with the support of small business funding from the National Institute of Child Health and Human Development (NICHD) and the National Institute on Aging (NIA) of the National Institutes of Health (NIH).
  • Slide 4
  • Homepages
  • Slide 5
  • Setting the Context As we grow, we learn to make sense of the world through inductive and deductive processes: Generalizing from observations to concepts and relationships; Then seeing and experiencing the world in terms of those generalizations. These processes can be enabling or disabling.
  • Slide 6
  • Makes a Man Proud to Be a Frog Many years ago, Walt Kelly offered this punch- line to an old nursery rhyme: Hi diddle, diddle; The cat and the fiddle; The dish ran away with the spoon; The little dog laughed to see such sport; And the cow jumped over the moon. And the three frogs stood at attention, saluted, and said: It makes a man proud to be a frog.
  • Slide 7
  • Seeing Is Believing and Believing Is Seeing More recently, Agnes, the creation of T. Cochran and a precocious little girl who can pontificate on any topic, responds to her friends suggestion that a little research might be appropriate with: Research is for the faithless.
  • Slide 8
  • Access to Data Is Not Enough Intelligent use of data requires that users: Know data and metadata; Know how to manage and analyze data; Know how to interpret results; Know how to apply results to problems; Know the limitations of their data, tools, and selves, This knowledge comes through experience and the understanding that there are few simple answers to lifes problems.
  • Slide 9
  • Making Sense of Data For inexperienced users of complex data sets, mastering the data and metadata can be a formidable task: Which data sets are relevant to a given concern? How good are the data? Are the items acceptable measures of more general concepts? What are the nuances and gotchas? Our objective is to speed and augment the acquisition of the experience needed to make informed and wise use of data.
  • Slide 10
  • Introduction to PDQ-Expert Digital representations of data make it possible to have intelligent interactive social science data capable of helping users formulate questions, specify analyses, and interpret results. This presentation describes our progress on developing an intelligent user interface for the PDQ-Explore information system that strives to achieve some of these objectives. The interface uses the Idea Works Qualrus Intelligent Qualitative Analysis Program to imbed case-based reasoning within a system of logical rules and semantic networks.
  • Slide 11
  • Qualrus Qualrus provides tools for creating a semantic network: Concepts, operations, and empirical measures linked in a network by logical rules. These are used with case-based reasoning to point the user to relevant examples. The examples serve as starting points for analysis and re-analysis.
  • Slide 12
  • Qualrus Link View
  • Slide 13
  • Qualrus Code Editor
  • Slide 14
  • Qualrus at Work Where do women earn the most money?
  • Slide 15
  • Qualrus at Work: Encore Are whites better off than blacks?
  • Slide 16
  • PDQ-Explore The PDQ-Explore information system combines paralleled high performance processors, data cached in random access memory, and efficient retrieval algorithms to process, in effect, tens or even hundreds of millions of records per second. Complex queries can be defined and executed in real time to produce tabulations, summary statistics, correlation matrices, and data extracts.
  • Slide 17
  • PDQ-Explore at Work--Setup Where do women earn the most money?
  • Slide 18
  • PDQ-Explore at Work--Results Where do women earn the most money?
  • Slide 19
  • PDQ-Explore at Work Are whites better off than blacks?
  • Slide 20
  • PDQ-Explore Quality of Housing
  • Slide 21
  • PDQ-Explore at Work State-to-State Migration
  • Slide 22
  • The PDQ-Expert WWW Interface The WWW Interface for PDQ-Expert lets users type in a free-form question based on U.S. Census Microdata (IPUMS). Users enter their question in the Query field then click on the Submit Query button.
  • Slide 23
  • The Systems Understanding of Your Query The first part of PDQ-Experts response to the user is a restatement of the users question as understood by the system. The purpose of this is to help the user see what the program thinks they are asking so they can identify any areas where the program may be going astray.
  • Slide 24
  • Refining Your Query After reviewing what the program thinks the user was asking, the next step is to consider key concepts from their question they may want to change or clarify. For example, here the original question suggests the concepts, sex, total personal income, and 1990.
  • Slide 25
  • Modifying the Query to Address Related Concepts Clicking the Hide/Show button associated with a concept shows a list of that concept along with other similar concepts that the user may want to examine. Users can check additional related concepts or substitute them for the original by unchecking it.
  • Slide 26
  • Similar Cases The next part of the feedback to the user shows a list of previous questions ordered with the ones most like the current displayed at the top of the list. Clicking the Hide/Show button displays details for that previous question. Users can incorporate elements of previous similar queries into the current one.
  • Slide 27
  • User Assessment and Saving of the Result in the Case File The final step in the CBR process is to assess the users satisfaction with the result, then save the resulting query along with the original question and all of its relevant parameters as a case in the database. Successful cases will be given scores that encourage them to be displayed in the future, while unsuccessful cases will be used to help avoid repeating past mistakes.
  • Slide 28
  • The Database of Previous Cases Each query, including both the text of the question and all codes associated with the question describing the resulting query, is saved in the database for consideration when the user enters future questions.
  • Slide 29
  • Summary This approach to helping users overcome their lack of knowledge about data and metadata appears to be a fruitful strategy that promises to provide a versatile and powerful interface to census and similar microdata. The approach has three specific strengths:
  • Slide 30
  • (1) Clarifying the Unknown Users (not just novice users) often do not know precisely what they want to ask, which data sets are relevant and appropriate to their concerns, and the characteristics of the items in the data sets. Users can be helped and their thinking clarified by viewing examples of similar questions and the queries they generate.
  • Slide 31
  • (2) Serving Diverse Users Novice and experienced users tend to ask very different kinds of questions and to need different kinds of help. This CBR system provides an effective way to supply diverse users with informative feedback and suggestions.
  • Slide 32
  • (3) Quick Implementation along with Continuing Improvement This CBR strategy can be put into place relatively quickly. It provides a framework for continued improvement in the knowledge of the system as new cases are added.
  • Slide 33
  • Current Status The development of PDQ-Expert has continued over the past year. Our focus has been on handling large semantic arrays more efficiently. We are now working on the link between the PDQ-Expert Interface and the PDQ- Explore interface and backend.
  • Slide 34
  • Thank You Albert Anderson, Public Data Queries, Inc. Edward Brent, Idea Works, Inc. Pawel Slusarz, Idea Works, Inc. Heather Branton, Public Data Queries, Inc. IASSIST 2003 Ottawa May 30, 2003