- Realization of Natural-Language Interfaces Using Lazy Functional Programming, Frost, 2006. I never ever heard of this article, with only 17 citations overall (in 5 years) it can hardly be considered important.
- In the entry of Transformation-based error-driven learning and natural language processing, Brill, 1995 (which is an important publication) it is stated that it "Describes a now commonly-used POS tagger based on transformation-based learning." Which is not true, since nearly everyone uses HMM, maxent, or SVM taggers these days because they give far higher accuracies.
Although it is far from perfect, the number of citations is probably one of the best manners to count importance. Someone actually did this per year for ACL conferences:
Obviously, there are other conferences, journals, etc. But it gives a pretty good overview of papers that are recommended. Also, there's the ACL top-10 rankings:
Those who don't learn from the past are destined to reinvent it, poorly.
"Latest and greatest" is not biased, in theory, but in practice it can be a lot harder to identify what truly is the latest and greatest as opposed to what is merely the most popular.
He talks about the little optimizations you can do in your life to take your research to world-class level.