The representation is in two parts: the "time series" itself and
the "interaction context", both of which are specified in the
project # 2 page. These two items have not been discussed
at length, so let me offer 2 examples below.

First, when you think of "time series", think: syntax - you are
representing the structure of the time series itself using 3D
symbols that bear no specific relation (picture-wise) to
the semantics carried by the time series. Think about 2D
diagrams - do they look like the thing that they are capturing?
Line, bar and pie charts do not look like the items that they are
representing. Lines, bars, and pie segments are acting as
"symbols".

The "interaction context" is for the semantics - what the time
series actually represents "meaning-wise".

.................

Here are 2 hypothetical news articles, fake but to get across
the above differences:

1. News story:  Scientists have become alarmed at recent solar
   flare activity over the past 20 months.

   First, you have to show the time series. Don't try to show it by
   showing 3D images of the sun, or spheres that look like the sun.
   Instead, take some attribute of the sun (perhaps the solar flares, or the
   notion that the sun is yellow and use these characteristics). For
   example,  show 20 flare-like particle systems of differing height
   shooting out of urns. The height of the particle system maps to the
   level of solar flare activity.

   The interaction context, though, is where you might show a 3D yellow sphere
   with solar flare activity - a kind of scientific vizualization.  Maybe, the
   time series flares form a straight line with the interaction context at the end,
   or perhaps the flares form a circle and the sphere hovers in the middle.

   The vertical flares are syntax for the time series, and the thing that actually
   looks like the sun, with its flares, is the semantics.

2  News story: People are starving because of food shortages in Country X
   over the past 3 years.

   First, you need to show a time series - now here's the thing - with something
   like # of people starving, this is harder to represent concretely (its like #
   of votes, # of bills, stock market values)--they are concepts. And so one is
   not tempted to show the time series with actual representations because there
   aren't any. Maybe, you simply decide to take some food item (like bread) and show
   a single large bread loaf growing and shrinking according to the starvation value.

   For interaction context, you have to ask yourself what the "meaning" is for
   the time series. Perhaps, showing someone who is sad or crying because they
   have no food for their family.


I hope that these 2 examples help a little:

The time series is a syntactic representation - there may be some semantic characteristics
and attributes but the goal is to create something more symbolic (and not directly
representational of what is happening over time).

The interaction context, on the other hand, does represent (with more meaning) what
is happening over time, and is more representational -- i.e., the semantics of the time series.

You may wonder what is the point of having a symbolic representation -- to show people
starving over time, why not just show this directly? The answer is that symbolism can be a powerful
way to convey information without us getting lost in visual or auditory details of what may
be more associated with the context.