6.835  Intelligent Multimodal User Interfaces

Please see the course schedule on stellar for the revised description of class next Thursday 5/7. Instead of optional office hours, I will summarize the lectures we didn't have in order to make room for the in-class studio sessions, focusing mostly on one really interesting issue: what do we mean by meaning, and how do we learn meaning, eg of a new word? There's an optional reading posted, no questions to answer.

Presentations and Final Projects: Spring 15

·         Projects, papers and slides from your final presentation are due at 6PM on Friday May 15^th . Turn in all of them in the usual way on Stellar.

·         Projects should be in in a zip file a form that is runnable by us if at all possible, i.e., unless they require specialized hardware other than a Leap or Kinect. In that unlikely case, please consult with us about what to turn in. We’ll work with you to get something reasonable that we can still try out. In all cases be sure to include a README file with instructions for setting up and running your code.

·         A crucial thing about the project is what you learned from it, not whether it worked perfectly. Good performance that you can’t explain isn’t worth much, while disappointing performance that arises for an interesting reason can be quite important.

·         Content: the paper should cover at least the following items

o   What is the task? Motivate it – why is it interesting?

o   Describe the system – what does it do? Show a real example of input and output.

o   Describe how it works. Note that this should be at the conceptual and architectural level, not at the level of code. Describe what ideas, techniques, and tools you used (e.g., for speech understanding, generation., etc.)

o   Describe how well it works. What experiments did you do? What did it do well and what did it do badly? Provide examples of both outstanding performance and underwhelming performance. Do the best you can to explain why it had this performance.

o   What in the implementation turned out to be more difficult than expected and why? (e.g., the Leap sensor was insufficiently accurate in some way, which you should describe). What worked easily?

o   Describe what you would do next to improve the system, to take its performance to the next level.

·         Logistics

o   Papers should also be turned in by hardcopy by the same deadline, in Prof. Davis’s office (32-237). We ask you to print it because then each of you does 10 minutes worth of work, rather than the course staff a few hours’ worth of work, plus this avoids the inevitable difficulty of getting fonts and/or figures to print properly.

o   Expected length is 10-15 pages, covering the material listed. Papers will be judged on content, insight, etc., not length. If you can be brief and substantive, fine; just make sure you address the issues above. Conversely, 20 rambling pages will not be well received.

o   Papers are individual work. Collaboration on the project should be indicated in the write-up: indicate who did what in building and testing the system. But you are expected to do independent write-ups. There will be obvious overlap in content, given that you are describing the same thing, but you should not subdivide the write-up work and each do only part.

·         The basic content for both the presentation and the papers is the same. Obviously the talk will cover these briefly, while the paper is expected to go into more detail.

·         Given the opportunities you’ve already had for presentations this term, the final presentations should be focused. They should remind people about the task and motivation, and provide a real example, but should focuson what you’ve learned, and any surprises/insights about performance and implementation (as above). A live demo is fine if you want to, but is not required; screenshots or a video are also ok. Explain what you learned, what came out of your user study, etc.

·         As noted, after your talk please turn in on Stellar a copy of your slides.

·         You must turn in any borrowed hardware to Professor Davis, Jeremy, or the course secretary in person. Replacing these is time consuming and expensive, so we will unabashedly hold your grade hostage until we get your hardware back.