Upload
alexandra-wells
View
212
Download
0
Embed Size (px)
Citation preview
1
What do People Recall about their Documents? Implications for Desktop
Search Tools
Tristan Blanc-Brude and Dominique L. ScapinINRIAACM IUI 2007 (22%)
2
Objective
Find out what users remember about their own documents
Provide suggestions for improving the design of personal file search engines
3
Participants
Seven researchers Seven members of administrative staff
4
Procedure
Participants were interviewed about their work, environment and how they manage their files
Participants talked about the workflow of their usual job tasks
30 documents belonging to each participant were selected by the experimenter based on task description
– 9 physical, 21 electronic Documents include ones that the that were
– consulted, modified or created by the participant– old, recent or recurring
5
Two Phases
Recall Phase – determine what participants remember about their files
Retrieval Phase – find out how participants search for personal files
6
Recall Phase
Free-recall – let participants talk freely about what they remember about the files
Cued-recall – give participants a list of file attributes and ask them to give the values
7
Recall Phase Attribute List
Location – complete file path Type or Format – file type (e.g., PowerPoint) File Name Title – title of document inside file Size – number of pages, lines, not bytes Time – last usage Keywords – inside document Visual elements – graphics, tables, colors inside document Associated events – related to the last document usage (e.g., emails, t
elephone calls) Links – other related documents Actions – operations performed on the document (e.g., printing, inserti
on, …, etc.)
8
Result: Frequency of Free-Recall Attributes
Textual Content – 71.4% Visual Elements – 25% File type or format – 21.4%
9
Result: Frequency of Cued-Recall Attributes
Recalled, but not necessarily correct
10
Result: Correctness of Recalled Attributes
Partial Recall: A part but not all of the description is correct. Time and size given in ranges.
Associated event, links and actions are left out, because they are hard to verify
11
Result: Frequently-used Descriptions
Size – approximation and ranges (e.g. about 50 pages, between 10 to 20 pages)
Time – periods (e.g. in February 2007) Visual elements – color, format and layout
(e.g. double column), presence of objects other than text (e.g. pictures)
12
Retrieval Phase
Main search method used by almost all participants– Browsing through folder hierarchy– Find the appropriate file in the folder
13
Result: Retrieval Phase
5 participants failed to find at least one document
The hardest part is finding the correct file folder
14
Attribute Expression Recommendation
Time – in period Size – approximation or interval Type – higher level category (e.g., slides inst
ead of ppt)
15
Attribute Specification Flexibility Recommendation
Wording – users can recall both correct and erroneous words
Size – approximation and interval Time – variable time units (e.g. year, month,
weeks)
16
Search Results Extensibility Recommendation
Location – show files in subdirectories in case the user does not remember the correct directory
Time – consider nearby dates in case the user does not remember the exact time
Size – error tolerance of +- 35%
17
Content Visualization Recommendation
Show previews of files other than the first pages, because first pages of similar documents can be very similar
18
Document Relationship Recommendation
Show the relationship between documents (e.g. copy/paste)
19
Operation Log Recommendation
Show a list of operations that had been done a document
20
Conclusion
Make the easily-recalled attributes available to users as search options