Docstoc

df

Document Sample
df Powered By Docstoc
					   Facilitate Humanities Researchers in
Sensemaking and Information Seeking Tasks




                     Yan Qu
               School of Information
                 yqu@umich.edu
               Project Overview
• What‟s the research problem?
   – How could a software system facilitate humanities researchers in
     their sensemaking and information seeking tasks?
• Why it‟s important?
   – Sensemaking and information seeking activities are ubiquitous in
     our research activities. The better support for these activities
     could result in efficient and high quality research work.
• What has been done?
   – Various sensemaking and information-seeking related theories
   – Existing systems
• What is my approach?
   – Study of sensemaking and infomration seeking behavior
   – System design and implementation
   – System evaluation
              Presentation Outline

• Introduction to the research problem
• Study of sensemaking and information seeking
  behaviors with the emphasis on sensemaking
  tasks conducted by humanities researchers.
• How such study informs the system design
• Related systems
• System design
  – System framework
  – Functions support sensemaking and information gathering
• System evaluation
            Research Problem

• How could a software system facilitate
  humanities researchers in their sensemaking
  and information seeking tasks
         Research Methodology
• Study of sensemaking and information seeking behaviors with the
  emphasis on sensemaking tasks conducted by humanities
  researchers.
   – Review general theory about sensemaking and information seeking
   – Investigate how humanities researchers conduct sensemaking and
     information seeking activities
   – Review theory about academic reading and writing
• How such study informs the design of software systems that support
  sensemaking and information seeking activities
• System design and implementation
• System evaluation
    Theories of Sensemaking and
        Information Seeking
•   Dervin‟s sensemaking model
•   Russell et al‟s sensemaking model
•   Information seeking model
•   Integrated model of sensemaking and
    information seeking
        Dervin‟s Sensemaking
                Model
•   Dervin‟s sensemaking triangle show the close relationship
    between sensemaking and information seeking activities


                           Situation



                  Help                    Gap
                 Russell et al‟s
               Sensemaking Model
•   Russell‟s sensemaking model posits the use of representations in service
    of accomplishing some tasks.


                 Search for good
                  representation              Generation Loop


                                               Representational
     Representaaions                Residue    Shift Loop


                   Instantiate
                 Repersentations               Data Coverage Loop
      Information Seeking Cycle
•   Information seeking is a rich heterogeneous and iterative process.
    [Bates 1989] [Pirolli and Card, 1995]

                              Expressed
                          Informtion Needs
         Formalize                                  Search


    Information Needs                                 Information
                                                         World
          Understand
           and Use                                  Search
                           Search Results
       Combined Model of Sensemaking
          and Information Seeking
•   Sensemaking and information seeking are always happened together
    and interweave with each other
•   Representation plays a central role in this model

                  Residue,
              Knowledge gap                       Expressed
            (Information Need)                Information Needs


                                                                  Information
        Representational   Representation                            World
           Shift Loop


                                               Search Results


                           Representational
                              Evolution
     Design Implication (Part 1)
• Support Sensemaking
  – Support storage and editing of multiple
    representations
  – Give suggestions on organizing the collected
    Information
  – Show different facets of the information
  – View representation at different granularities
• Support Information Gathering
  – Help express the information needs
  – Organize search results
     How Humanities Researchers Conduct Their
    Sensemaking and Information Seeking Tasks (1)

•       The study of “The Night before Christmas” -- Stephen
        Nissenbaum
    –     Study of the role the poem played in historical social change.
    –     Moore‟s poem was about the privatization of Christmas. And
          the poem itself seems to have played an important role in
          bringing about that change.

•       What we see in this sample of humanities research
    –     The representation of knowledge is gradually evolved.
    –     Search in wide range of information resources.
    –     Various activities are involved: information selection,
          categorization, connection, reasoning, comparison, etc..
  How Humanities Researchers Conduct Their
 Sensemaking and Information Seeking Tasks (2)

• Research methodology
  – Qualitative “data”/observation, qualitative reasoning, non-
    experimental
• Research Process/Activities
  – Gathering data (mostly qualitative, primary and secondary
    sources)
  – Process/understand data: organize, interpret, qualitatively
    reasoning, etc.
  – Writing and publishing
• Features:
  – Wide variety of data resources
  – Intensive information related activities: organization, connection,
    aggregation, comparison, reasoning, and interpretation.
           Design Decision
• Mainly focus on support of sensemaking
  activities
• Support organizing collected information
• Support reading and writing
• Scenario that resembles typical
  sensemaking and information seeking
  actives in humanities research: a
  researcher constructs a paper outline.
 Reading as a Sensemaking Activity
• “Readers do not „receive‟ information. They approach
  reading in the context of the entire world of their
  experience, and they turn away with that world
  confirmed, modified, extended, or challenged” [Gerald,
  1994]
• “Reading is thinking stimulated by print” [Vaughn 1984],
  it includes various of activities: categorization;
  comparing, connecting and organizing ideas; clarifying,
  generating questions; analyzing, synthesizing… [Gerald,
  1994]
• People‟s writing are representations of their knowledge.
  Reading is a process transferring other knowledge
  representation to our own.
 Writing as Construction [Bolter, 2001]

• A writer begin with a jumble of verbal ideas and only a
  vague sense of how these ideas will fit together.

• He may start by laying out topics in an arrangement less
  formal than an outline: he may organize by association
  rather that strict subordination.

• Then the writer trims his network by removing
  connections and establishing subordination until there is
  a strict hierarchy.
      Design Implication (Part 2)

• Facilitate reading comprehension
  – Help note taking
  – Give suggestions on organizing the collected
    Information (same as part 1)
  – Show different facets of the information (same as part
    1)
• Support outline generation
  – Support brainstorming and idea formation
  – Help representation shift from network to tree
    representation
      Design Implication (Repeat)
•   Support storage and editing of multiple representations
•   Give suggestions on organizing the collected Information
•   Show different facets of the information
•   Help note taking
•   View representation at different granularity
•   Support brain storming and idea formation
•   Help representation shift from network to tree representation
•   Help express the information needs
•   Organize search results
•   Support the iterative information seeking process
                  Related systems
• Research systems:
   –   NoteCrads [Halasz, Moran, and Trigg, 1987]
   –   gIBIS [Conklin and Begeman, 1998]
   –   SenseMaker [Bell, 1997]
   –   SenseMaker [Baldonado and Winograd, 1997]
   –   PowerBookmarks [Li, et al, 1999]
• Brainstorming software:
   – MindManager
   – ThoughtPath
• Bookmark manager:
   – Check&Get
   – Private Bookmarks
       My Approach Differs in:
• A software system is designed to facilitate
  humanities researchers in their sensemaking and
  information seeking tasks.
• I take as central to my approach the representation
  search and shift in sensemaking;
• Machine learning techniques are intensively used to
  provide automatic features in information
  organization and presentation structure
  manipulation.
                 System Design - Interface
  Tree of Gathered
    Information

                                      Network/Outline
                                       representation




Tree Node Summary
       System Design - Structure
                    User Interface




Network structure              Text Processing
 Manipulation                                               DB
                                                  Users’ Representation Info
                                Clustering and
 Tree Structure                 Classification         Web Resources
 Manipulation
                                Outline Editing




             Vivisimo Clustering
               Search Engine



                     Web
Features to Support Sensemaking

• Support Sensemaking
  – Support storage and editing of network and tree
    representations
  – Give suggestions on organizing the collected Information
  – Show different facets of the information
  – Note taking functionalities
  – Help generate the outline
System Function – Basic Tree
          Editing


                                  Tree Structure Editing



                                  Tree Node Editing




(a) Interface of basic tree editing tools                  (b) Interface of keywords editor
System Function – Basic Network
           Editing

                                                 Network Structure
                                                 Editing Pane




                                                 Gathered data is organized
                                                 in different groups




      Interface of basic network editing tools
                               System Function –
                             Clustering/Classification
                        Clustering documents
                        under the selected folder




                 Classify documents under
                 the selected folder                (b) Example clustering result   (c) Example classification result with the sample
                                                                                    documents marked with different background

(a) Interface of Clustering/Classification Tool
    Clustering/Classification
          Algorithms
• Feature Selection
  – Document frequency thresholding (DF)
  – Information gain(IG)
• Clustering algorithm - Single-linkage
  clustering algorithm
• Classification algorithm - Semi-
  Supervised KNN
   System Function – Tree Structure
            Comparison
• Two sub-trees from different category schemes are
  selected, and a node from one sub-tree is colored if it is also
  appears in another sub-tree.




 (a) Interface of tree structure comparison tool   (b) An example of tree structure comparison: the
                                                     distribution of economic news in China news
System Function – Note Taking




    Interface of note editing tool
               System Function – Search
                                         Choices of how
                                         the query forms




                                         Choices of how the
                                         result is organized




(a) Interface of the basic search tool                         (b) Search result organized in tree structure
      System Functions Under
           Development
• Outline editing tools
• Support transformation from network
  representation to outline
• Export outline into XML format
                 System Evaluation
• Qualitative test
   – Study how doctoral student in humanities using this
     system in their research.
      • Observation + Interview
      • Find out if the system is helpful in those student‟s research
      • Find out the advantage and disadvantage of using this system,
        comparing to how the students do the tasks before.


• Quantitative test
   – Study whether users who use this system come with
     better sensemaking results than users who don‟t use
     this system.
      • How to define “better performance” in sensemaking tasks?
Questions




?

				
DOCUMENT INFO