Dear friends,
Action item: Michal
MÄchura will look at the schemas of existing dictinaries
(those available through Elexis, and others) and try to come
with some emprical observations on recursion (and
recursion-like things): how prevalent is it, how many levels,
what phenomena is it used to model, could those phenomena be
modelled using something else, what would be lost or gained in
such remodelling? Michal to submit this analysis in time for
reactions before next meeting..
I havenât had enough
time to do this action item properly yet, as you can see from
the fact that I havenât actually sent anything. But I did take a
preliminary, informal look at a few entries in a few German,
English, Irish and Czech dictionaries, and these are my initial
impressions regarding recursion/embeding on senses:
-
I found it more informative to look at how
dictionaries are presented on screen (on the web) rather
than reading their XML (or whatever) source code.
-
It seems that the more âmodernâ or ârecentâ or
âborn-digitalâ a dictionary is (which is difficult to
define, I know, but bear with me), the less
likely it is to use sense-inside-sense embedding at all.
And, if a dictionary does use some form of sense embedding,
then the embedding appears to be more formally restricted if
the dictionary is relatively âmodernâ. By ârestrictedâ I
mean that (for example) the recursion never goes deeper than
one level down, or that senses are grouped into
âsupersensesâ by part of speech but the âsupersenseâ
contains no additional data beside the part-of-speech label.
-
So far I havenât seen a single instance of
sense embedding which couldnât be modelled by some other
means, such as by the methods mentioned in my notes from our
last meeting (sense-to-sense relations, or tags/labels), and
where that modelling could not be done automatically
(without human intervention). In other words, I havenât seen
a single instance of embedding where the sense list couldnât
be flattened to just a single level without any noticeable
loss of information or understandability or usefulness to
the reader/user. In yet other words, I havenât seen a single
instance of embedding where the child sense would stop
making sense (pardon the pun) if it became disconnected from
its mother sense. But, that said, it is possible that if I
keep looking I will eventually find some counter-examples â
so, consider this a hypothesis, not a conclusion.
I will do my best to
write this all up more fully, and to illustrate my points with
specific examples, in a more technical report which I will send
along by the next meeting. Sorry this is taking me longer than
expected but, you know, this is not my day jobâ :-)
Also, it is possible
(though not certain) that I will not be able to come to our
meeting later on today, due to some conflicing appointments. If
I donât turn up please accept my apologies and do not wait for
me.
Thanks you,
Michal
â
|