8:00 – 11:00 SR
- Created a timer callback, modified the dataprovider and re-rendered the chart. Seems to be working fine, with no flicker, even with a 100Msec update rate.
11:00 – 1:00 FP
- Meeting with Andrea. Looks like uniqueness can be determined by word choice in a corpus as small as 500. That does make things easier. It also allows for triangulation against other metrics, which would allow for looking at accelerometer data from several body positions for example. Though I’m not sure that’s needed.
- An issue to consider is that people who might use the site would have other examples of their writing. This means that an anonymous source could be identified. As a way around this, a vector-based translation algorithm could be trained using identification code to remove/modify the parts of the user’s language that are identifiable.
- And actually, this means that I could build a simple website that you train once, then write to. The site then transpiles the user’s words and publishes to twitter or Facebook for example. This just addresses the anonymous part of the problem, not the trustworthiness part, but it’s nice low-hanging fruit.
1:00 – 4:00
- Added a datatable, which behaves better. For example, the set() method fires a refresh. Spent the rest of the day trying to get the same behavior with the chart and finally wound up posting on the forums.