Theiling Online    Sitemap    Conlang Mailing List HQ   

Thoughts on Word building

From:Gary Shannon <fiziwig@...>
Date:Saturday, December 3, 2005, 21:51
Prefixes and suffixes added to a root word can be used
to derive more words in languages structed that way.
But what is the "best" root, and the "best" set of
derivations for any given concept. The derivation can
proceed in any direction, but there must be some
particular root or set of roots that results in an
optimal tree with the shortest or most understandable
derivatives. For example, given a set of words having
to do with information: to know, knowledge, known,
knowledgable, to teach, to learn, teacher, student, to
forget, to remember, lesson, ignorance, scholar,
dunce, etc. ANY ONE of those words can be used as the
root from which all the others can be derived.

know -> know-stuff (knowledge) -> know-stuff-give (to
teach) -> know-stuff-give-person (teacher).
teacher -> teacher-job (to teach) -> teacher-job-stuff
(knowledge) -> treacher-job-stuff-have (to know).
ignorance -> ignorance-remove (to teach) ->
ignorance-remove-person (teacher) ->
ignorance-remove-person-client (student).

If all these words were arranged in an interconnected
multi-dimensional network, where the paths linking
adjacent words (nodes) represented the meaning of the
prefix or suffix connecting them, then there cannot be
such a thing as a "most primative" word. Any word can
be taken to be the most primative word and all other
can be shown to be derived from it.

So the question is not what words are more primative,
but rather, what distribution of arbitrary root words
in the network result in the "best" set of derived
words? Here, "best" will have to be defined according
to the design goals of the language.

So the first question is what is the optimum set of
affix pairs? They will be pairs because they must be
bi-directional as in doer/job (to_teach + doer ->
teacher; teacher + job -> to_teach) so that each
member of the pair un-does the other member. (doer +
job = NULL, so that to_teach + doer + job = to_teach).
How many affixes exist in English? There must be a
bunch of them. In five minutes, just off the top of my
head I have: (The letters in brackets are replaced by
the suffix)

-[ce]-tific science -> scientific
-[y]-ic geology -> geologic | history -> historic
-ic[]-al geologic -> geological | tropic -> tropical
-[os]-ic cosmos -> cosmic
-[o]-ic volcano -> volcanic
-n[]-ic titan -> titanic | electron -> electronic
-[an]-c barbarian -> barbaric
-[an]-ism barbarian -> barbarism
-[]-agoric phantasm -> phantasmagoric
-[e]-ic -> automate -> automatic
-[y]-iance comply -> compliance compliance | vary ->
-[]-ance -> appear -> appearance | accept ->
-[]-ence correspond -> correspondence
-[ect]-igence neglect -> negligence
-t[]-ion invent -> invention
-[y]-ial deny -> denial | try -> trial
-[y]-ful beauty -> beautiful | pity -> pitiful |
plenty -> plentiful
-t[e]-ion obligate -> obligation | automate ->
plus -able, -ible, -ment, -er (doer), -ive, -ative,

And on and on. It seems like it would be very handy to
have a systematic list of such derivational
components. Even things like being able to derive
"pizzaria" from "pizza" and "happy" from "sad" makes
the job of vocabulary building much simpler.

So anyway, the point of all this rambling is; it seem
like a very good starting point for a conlang
(assuming it is structed to be able to use prefixes
and suffixes) is to collect a comprehensive set of
conlang affixes and compounding rules. After that, one
single primative root can yeild dozens, or maybe
hundreds of additional words by affixing and

But is there in existence on the web such a list of
affix functions?



Larry Sulky <larrysulky@...>
Aaron Grahn <aaron@...>
Aaron Grahn <aaron@...>
Eduard Ralph <conlang@...>