22. Quantitative approach
• Categorization of user requirements
• According to different user groups
• Prioritize implementation of system features
1. 2. 3.
24. Method
• Phase 1: “Idea generation”
–Caveat: unfamiliarity with A/V search
–Interviews, requirements elicitation
experiments using system mock-ups
–professionals, researchers, home user
–71 “statements”
25. Method (2)
• Phase 2: “Sorting & Rating”
–47 participants, evenly spread over 3 groups
–Categorize according to perceived similarity
• Create categories and label with description
–Rate according to desirability
• 1-5: undesirable – highly desirable
28. User-defined search functions
Translates into:
1. having options for search manipulation and
result representation:
– change sort order of results (date, relevance)
– Use filters and activate them at any point during
search
– Use boolean operators
2. the system should be clear and guide users in
a rather complex search process
29. Advanced search
• Users are eager to make use of advanced search
options:
– Search using example image
– Events, persons, speakers, visual objects
– Using multiple metadata sources, e.g.,
• Archival metadata
• Speech transcripts
• Visual labels
• Less interest in “technical” metadata
– Shot types, camera movement, b/w-color
30. Help functions
• Slight miss-alignment with what developers
understand as “help functions”
• Next to “access to help manuals”:
– System should be fast
– Clear, user-friendly, intuitive UI
– Access to video
– Clear message when no results are found
– Consistent results
37. Conclusions
• Quantitative approach using concept mapping
– Useful approach to get some more grip on user
requirements and differences between user
groups
– Helps to steer development process of advanced
user type specific A/V search tools
– User defined labels may not always be in-line with
formal definitions used by developers which may
cause confusions, e.g.,
• Help functions vs. clear, smooth and consistent system
38. Conclusions (2)
• Clear interest of users in
1. advanced search options
2. segment-level access points to AV by means of
rich multi-modal annotations (speech, speakers,
events, objects)
3. Clear, transparent, smoothly working system