============================================================================ REVIEWER #1 ============================================================================ --------------------------------------------------------------------------- Reviewer's Scores --------------------------------------------------------------------------- Appropriateness: 1 Clarity: 2 Originality /Significance: 2 Soundness / Correctness: 2 Recommendation: 2 Reviewer's Confidence: 2 --------------------------------------------------------------------------- Comments --------------------------------------------------------------------------- The authors present an interaction technique for icon-based AAC systems, which does not rely on discrete one-by-one selection. Rather, the user draws a continuous path over the two-dimensional array of icons, containing the ones he or she wishes to select. The system then determines the most likely subset among the icons on the path. At first, I thought this is nothing else but the application of Swype - the predictive text entry technique for touchscreen-based mobile phones (en.wikipedia.org/wiki/Swype) - to general icons (as opposed to letter-icons only). Therefore, I was surprised that Swype is not even mentioned once in the entire paper. However, the system also reorders the selected icons, in order to produce syntactically correct sentences/utterances. Reordering is something Swype does not do (although I am not quite sure SymbolPath should do it, especially in the context of more complicated sentences). I think that the semantic and syntactic considerations involved are very interesting and that this work will make a very nice demo. ============================================================================ REVIEWER #2 ============================================================================ --------------------------------------------------------------------------- Reviewer's Scores --------------------------------------------------------------------------- Appropriateness: 1 Clarity: 2 Originality /Significance: 3 Soundness / Correctness: 2 Recommendation: 2 Reviewer's Confidence: 3 --------------------------------------------------------------------------- Comments --------------------------------------------------------------------------- This work addresses an interesting problem with AAC input and is well motivated. It would be a good topic for discussion during the poster session and a reasonable subset of the ASSETS community would find it interesting. A few questions / comments I have are: - How does the severity of hand tremors impact recognition accuracy? Have you tested it with users yet? - Can people maintain constant contact with the display or do they accidentally lift their hand during utterance construction? - When analyzing the path characteristics, do you factor in time in terms of pauses on an icon or speed of movement? - You could cite other literature on or examples of input techniques that are related (e.g., swipe text entry). ============================================================================ REVIEWER #3 ============================================================================ --------------------------------------------------------------------------- Reviewer's Scores --------------------------------------------------------------------------- Appropriateness: 1 Clarity: 1 Originality /Significance: 2 Soundness / Correctness: 2 Recommendation: 2 Reviewer's Confidence: 1 --------------------------------------------------------------------------- Comments --------------------------------------------------------------------------- This paper reports the design and development of SymbolPath, an AAC system that allows users to use continuous motion to select icons for communication. The motion path contains both the desired icons and the redundant icons. The system then uses a combination of semantic frames, semantic grams, and hysical characteristics of the motion path to generate a list of potential utterance. The work is very interesting and can potentially benefit a large number of users with both speech and physical disabilities. The weakness of the paper is the lack of user evaluations. It'll be helpful if some preliminary data is available by the time of the conference. I think the work will make an interesting demo at the conference.