Is there any plots, etc. that show performance metrics on various natural language processing tasks as a function of year.
For example the following figure shows the evolution of performance of face recognition systems:
I am looking for something similar for natural language processing tasks (e.g. evolution of the perplexity of the state-of-the-art language models, BLUE score of translation systems, accuracy of POS tagging, etc.).
The ACL wiki contains a section on the state-of-the-art systems over time for the main NLP tasks: http://aclweb.org/aclwiki/index.php?title=State_of_the_art
Many of these pages need to be updated but at list there is some attempt of collaborative work to construct a clear overview.
The only survey article the point to dates from 1997 (Survey of the state of the art in human language technology. Vol. 13. Cambridge: Cambridge University Press, 1997), so I fear there is none more recent.
Excerpt from one of the most satisfying page I found on the ACL wiki:
