Currently, maps tell us which topics are featured in a corpora but they do not give extra information on terms. Here are two dev direction that could greatly improve at low cost the information conveyed by the maps by means of colors. An implementation of these feature is given in http://risk.iscpif.fr/map2015/explorerjs.html?mode=db.json.

Notations

• A corpora with doc ranging from year $Y$ to year $(Y-T)$.
• Let $n_i^{[y_1,y_2]}$ be the number of documents of the corpora mentioning $i$ between years $y_1$ and $y_2$.

For corpora such that T>1, make the following coloring of the terms available :

1. Catch the trends (colors selector growth_rate in the example). Map the color of node on the dynamics of their occurrences. The dynamics of a term in the corpus is a nice indicator of the attention if gather.

• take t=min(3,N/2) (3 is abitrary but is the order of magnitude to the caracteristic time of science. A PhD for example takes 3 years.)
• compute for each term $i$, $n_i^{[Y-t,Y]}$ and $n_i^{[Y-2*t,Y-t]}$
• compute the ratio $r_i=\frac{n_i^{[Y-t,Y]}+1}{n_i^{[Y-2*t,Y-t] }+ 1}$ (we add one to avoid division by 0 ...)
• map the color of each node as follows :
  • if $0.95 \leq r_i \leq 1.05$, color = white
  • if $1.05 \leq r_i$, put a color map from yellow to dark red on [0 1] parameterized by $\frac{2}{\pi} arctan(r_i - 1)$
  • if $r_i \leq 0.95$, put a color map from green to dark blue on [0 1] parameterized by $\frac{2}{\pi} arctan(\frac{1}{r_i } - 1)$

2. Unfold the history. (colors selector age in the example) if Map the colors of a note according to relative time of appearance in the corpora as follow.

• for each term $i$, compute the first year $t$ of appearance in the corpora.
• if t=$Y-T$ then color the node as white
• else, map the color from dark blue to dark red passing to green and yellow according to it position on $\frac{t-(Y-T)}{Y-T}$

These two coloring option will help the user to have her eyes caught by interesting areas of the map in terms of dynamics.