Skip to main content
. 2006 Sep 26;7(Suppl 2):S3. doi: 10.1186/1471-2105-7-S2-S3

Figure 2.

Figure 2

MM training curves converge at different rates (light blue line = 200-period moving average). (A) MM training on non-scientific text – in this case, Tolstoy's "War and Peace". Note that convergence is faster and more stable than when trained on scientific text (B), which is more complex. (C) Training on chemical names requires a relatively large training set, but reaches convergence.