Article details

Research area
Speech recognition

Proceedings of the Applied Human Factors and Ergonomics Conference, San Francisco, CA, USA


Jan Curin, Jan Kleindienst, Martin Labsky, Tomas Macek, Hoi Young, Ann Thyme-Gobbel, Lars König, Holger Quast, Christophe Couvreur

Dictating and editing short texts while driving


Although several existing in-car systems support dictation, there is none which would systematically address dictation and error correction for automotive environments. Dictation and correction systems available for desktop and mobile are not suitable for the car environment where safety is the crucial aspect.

This paper presents a multimodal automotive dictation editor (codenamed ECOR), designed as a test bed for evaluation of numerous error correction techniques. Results are presented both for a standard use of the application as well as for the case when a particular type of correction is enforced. Reported results are obtained from native US-English speakers using the system while driving a standard lane-change-test (LCT) low fidelity car simulator. The dictation editor was tested in several modes including operations without any display, with a display showing the full edited text, and with limited view of just the “active” part of the dictated text. The measured results are compared to SMS dictation using a cell phone and to destination entry using a GPS unit.

The results indicate that the eyes-free version keeps the distraction level acceptable while achieving good task completion rate. Both multimodal versions caused more distraction than the eyes-free version and were comparable to the GPS entry task. By far, the cell phone texting task was the most distracting one. Text composition speed using dictation was faster than cell phone typing.

Read/download now