2. Composing Letters with a Simulated
Listening Typewriter
JOHN D. GOULD, JOHN CONTI and TODD HOVANYECZ
IBM Research Center
3. Writing vs Dictating
• Evaluation of a Simulated Listening Typewriter.
• 1983…
• Aim of the experiments was to determine if an imperfect
listening typewriter would be useful for composing letters.
5. The experiments
2 experiments
10 naive participants (10 assignments each) and 8 professionals (7
assignments each).
1000, 5000 and unlimited vocabulary.
Isolated and continuous speech.
2 strategies (draft and final).
Written letters were included.
6. Methods of the experiments
1. Observing (video logging)
2. Asking users (preferences, opinions, …)
3. Asking experts (evaluation of the produced letters)
4. Testing (time measuring, error counting, …)
7. Conclusion
• Some versions were as good as traditional methods of
handwriting and dictating.
• Isolated word speech + large vocabularies = a (potentially) useful
listening typewriter.
• Some participants (especially of the second group) had been
frustrated by the slow speed of the simulated listening
typewriter.
8. A critic view
• Speech recognition was under development (1983).
• What about “ehs…”, “ums…” etc?
• The text’s style (spacing, text orientation, …)?
• Complex and time spending instructions.
• Using of written notes was allowed (ambiguous ?).
• Evaluation for disabled people.