inference (vb)

Laurence Horn laurence.horn at YALE.EDU
Wed Jan 14 22:00:29 UTC 2009

Impressions from a longtime pragmaticist:
"inferencing" is standard usage for those working
on natural language inferencing (involving the
drawing of either semantic or pragmatic
inferences); "infer(r)er" is pretty rare, maybe
because it's hard to pronounce, but "inferencer"
is even less likely; "infer(r)able" is standard
usage, both as adjective and noun [= something
that is or is likely to be inferred], but
annoyingly results in a spell-check problem.
That is, while "inferrable" is intuitively the
appropriate form, spell-checkers perversely
insist on "inferable", which looks to many of us
(my colleagues and I have grumbled about this
issue) as though it should be pronounced with
primary stress on the first syllable (à la
"comparable" or "operable") instead of on the
antepenult, as it always is in fact pronounced.


At 3:33 PM -0500 1/14/09, Mark Mandel wrote:
>Not in OED, although the noun "inferencer", def. "One who draws an
>inference.", is listed with the flag "nonce-wd." The sole cite is from
>raw ghits (all specified with quotation marks):
>about 4,640 for "inferenced"
>about 216,000 for "inferencing"
>about 10,900 for "inferencer"
>about 1,190 for "inferencers"
>Current uses seem to be mostly in I.T. --
>1.  inferencers, which are algorithms for computing ( perhaps
>approximately) marginal distributions over nodes in the model
>2. This is my first experience with a type-inferenced imperative
>But not all. These are = the corresponding forms of "infer" AFAICT:
>3. Some adults are much better at inferring than others, but children
>are such natural inferencers that is hardly takes long until they
>thinking about their ...
>4. I decided to create a unit about inferencing because this is an
>area I struggled with last year and was not really pleased with how I
>pulled it all together ...
>( -- developing
>classroom lessons for elementary school teaching)
>Note that #3 has gerund "inferring" and agent noun "inferencer" in the
>same sentence, in direct contrast. This may be motivated by avoiding
>the /'@r at r/ sequence. So...
>about 97,800 for "infered", including a number of typoes for "infrared"
>about 16,400,000 for "inferred"
>about 42,200 for "infering"
>about 2,320,000 for "inferring"
>about 31,500 for "inferer"
>about 6,150 for "inferrer"
>65 for "inferers"
>    (A significant proportion of "inferer(s)" hits are typoes for
>about 971 for "inferrers"
>These are both I.T. and human, but my subjective impression is that
>"infer(r)er(s)" has a much higher proportion of human to IT references
>than does "inferencer(s)". That would be expectable if the latter form
>and/or the verb "inference" was used for early development of this
>technology: it would have spread quickly within the IT community,
>while the simpler form is probably (W.A.G.) formed productively by
>most of its users.
>5. Another possibility, for the recursive inferer, is to compute a ...
>The way the recursive inferer handles this is to keep two separate
>scores: a similarity ... (
>6. The inferer himself may have such doubt or some one else may raise
>it. ... So when the inferer has the cognition of the relation of
>vyapti he already knows ...
>As usual, the ghits should be taken with a tsp. of salt. F'rex, this
>one doesn't have the search word at all:
>7. Infering means to take what you know and make a guess. Read the
>following situations and pick which answer you could infer.
>Mark Mandel
>The American Dialect Society -

The American Dialect Society -

More information about the Ads-l mailing list