[Corpora-List] little favour
yorick at dcs.shef.ac.uk
Sun Sep 9 18:36:49 CEST 2007
Im sorry to get into this, but I dont think this reasoning makes very
What you describe are not experiments, surely? You COULD conduct an
experiment that asked a number
of people to make up sentences (using, say, "slip" as a verb) and you
could then ask "do any of the subjects make up the same
sentence?" That would be an experiment, but not an interesting one,
and it is surely not the issue here at all. If you
look in a corpus for sentences with that same criterion, then they,
again, may or may not be identical--they will almost certainly not be
but that is again irrelevant to the issue--but merely seeking
sentences, under any description, is again not an experiment.
The issue HERE is about whether or not spontaneous sentences that
serve a natural purpose , whatever it is, and which form a corpus,
are data OF THE SAME SORT
as sentences made up to serve the goal of being data--that surely is
the question, and it has nothing to do with whether the members of
the sets differ or not,
nor with any notion of experiment or replicability. Ramesh plainly
thinks they are not of the same sort, while others think they are, or
can be, or the difference doesnt matter
too much for what they want to do (Ulf seems to be in that group).
But surely THAT is the issue.
Sorry to sound so testy and dogmatic--it's been a slow day.
On 9 Sep 2007, at 16:51, Santos Diana wrote:
> I meant that, if you are asked to come up with 6 made-up sentences
> illustrating the verb slip one year after, you are going to come up
> 6 different examples. (I have no proof of this, though :-) I suggest
> that this is tried empirically...)
> So, you cannot replicate your experiment and get the same sentences,
> while if you ask that of a corpus you (or another researcher) would
> the same answer. I tried to point out this as one advantage of having
> corpora and not relying on made-up sentences by a (quite) unknown
> Of course, I am aware that once you _collected_ those 6 sentences from
> many people and create a corpus of made-up sentences, then you fix
> and can "repeat"/"replicate" your experiments.
> But while one knows that considerable work has been put in
> designing the
> BNC (and one knows what kinds of speakers and what kinds of texts it
> represents) -- and so you might think of creating a new/similar BNC
> later -- it is hard to know what kind of mood and what kind of person
> answered those 6 sentences, which is going to make it very unlikely
> you will be able to replicate this new "corpus".
> This is what I meant by the sentence below, sorry for the bad English.
>> -----Original Message-----
>> From: corpora-bounces at uib.no [mailto:corpora-bounces at uib.no]
>> On Behalf Of Mike Maxwell
>> Sent: 9. september 2007 17:08
>> To: corpora at uib.no
>> Subject: Re: [Corpora-List] little favour
>> Santos Diana wrote:
>>> Elicited made-up sentences are probably unrepeatable
>> Can you explain what you mean by this?
>> Mike Maxwell
>> maxwell at umiacs.umd.edu
>> "Theorists...have merely to lock themselves in a room
>> with a blackboard and coffee maker to conduct their business."
>> --Bruce A. Schumm, Deep Down Things
>> Corpora mailing list
>> Corpora at uib.no
> Corpora mailing list
> Corpora at uib.no
More information about the Corpora