Adventures discovering an English Affix Grammar


The goal of this project is to determine if english affixes (prefixes & suffixes), (or words more generally) can be stemmed into micro-affix forms through an automated, statistical process (hierarchical clustering and distance).  If so, we would like to determine both the classes the lexical alternations belong to, and their transformational characteristics.

Example: +ed
“ed” may be referred to as a time transformation, in that ed(C) takes action C, completes it, and moves it into the past. It also means that ed() can only be applied to proper X’s — C’s that have some temporal component (can happen in the past (for ed)).  When it is applied to an improper C, it’s transformational properties are reflexive — features are applied to C to make it proper.

For instance: C = “water”. “Water” is a noun, it cannot happen in the past. But let’s say that we ed(water). In making the word “watered”, we have stated something about “water”. We have applied the feature of “action, to do, to apply” to water, providing it the functionality of both a noun and a verb.

Another (+ed) example:

“He sure ‘Beracahed’ that!”
Beracah+ed = Applying the traits of, having an effect similar, or doing something like Beracah might.

+ed application to an unknown object allows us to conclude something about Beracah — besides being a noun, it has the ability to effect the world in a temporal manner.