[Lexicog] Semantic parsing

Ron Moe ron_moe at SIL.ORG
Wed May 31 21:53:53 UTC 2006


Patrick Hanks wrote:

I guess all this reinforces the point that semantic tagging is more
problematic than part-of-speeech tagging.

There are between 10 and 100 grammatical categories in a language. There are
between 1,000 and 10,000 semantic categories. So we are about two orders of
magnitude greater in semantics. I'm not a mathematician, but I believe
complexity works exponentially.

Ron Moe
  -----Original Message-----
  From: lexicographylist at yahoogroups.com
[mailto:lexicographylist at yahoogroups.com]On Behalf Of Patrick Hanks
  Sent: Wednesday, May 31, 2006 8:58 AM
  To: lexicographylist at yahoogroups.com
  Subject: Re: [Lexicog] Semantic parsing


  David Frank wrote:

  > What am I missing?

  Nothing  -- Well not much, anyway.

  The problem in this case is the word 'resemble', not FrameNet. You could
not have chosen a less instructive frame than the Similarity Frame in FN (of
which "resemble" is a member).  Anything can resemble anything, so semantic
tagging of valences isn't going to shed much light on the meaning.

  To be fair, this frame does mention that there is a [[Dimension]] in
respect of which similarity is measured. That seems somewhat helpful. (The
[[Dimension]] may be mentioned explicitly, or left to the imagination -- but
it's still there in the semantics of "resemble", even when not explicit.)

  FN says:

        Dimension []
       This FE marks constituents which express a property in respect to
which the similarity of the entities is assessed.
                 The disc announced Friday is physically SIMILAR to current
music CDs


  * *

  Comparison with "represent", which is in the Communicate_categorization
frame, is quite thought-provoking.

  I guess all this reinforces the point that semantic tagging is more
problematic than part-of-speeech tagging.

  Patrick


  ----- Original Message -----
    From: David Frank
    To: lexicographylist at yahoogroups.com
    Sent: Wednesday, May 31, 2006 5:01 PM
    Subject: Re: [Lexicog] Semantic parsing


    from David Frank:

    I was interested to hear about FrameNet. I am doing a study that
includes an analysis of the word "resemble," and I have been wondering how
Fillmore would analyze the case frame for that verb, as compared with, for
example, "represent," and I wasn't sure how to find out.

    However, after looking at the FrameNet web site, I am disappointed. The
analysis of "resemble" I found at
http://framenet.icsi.berkeley.edu/index.php?option=com_wrapper&Itemid=118&le
=7974& is something like the following (using my own example):

    That cloud resembles a bear = [[Entity_1]] resembles [[Entity_2]].

    Additionally, there is a definition given for "resemble": "have a
similar appearance to or features in common with." "Entity" as a frame is
defined as "A thing (either abstract or physical) that exists with some
degree of permanence." There are no insights as to the relationship between
Entity_1 and Entity_2.

    To me this is not very helpful. What am I missing?


    ----- Original Message -----
    From: <rtroike at email.arizona.edu>
    To: <lexicographylist at yahoogroups.com>
    Sent: Wednesday, May 31, 2006 4:01 AM
    Subject: [Lexicog] Semantic parsing


    Patrick's excellent and very clear exposition of the problem of semantic
    parsing is exactly what is being addressed in Charles Fillmore's
seemingly
    little known Framenet project, which is using frame semantics to set up
an
    automatic semantic annotation system. The url for the project is:

            http://framenet.icsi.berkeley.edu/


    SPONSORED LINKS Science kits  Science education  Science kit for kid
          Cognitive science  Science education supply  My first science kit


----------------------------------------------------------------------------
    YAHOO! GROUPS LINKS

      a..  Visit your group "lexicographylist" on the web.

      b..  To unsubscribe from this group, send an email to:
       lexicographylist-unsubscribe at yahoogroups.com

      c..  Your use of Yahoo! Groups is subject to the Yahoo! Terms of
Service.


----------------------------------------------------------------------------

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lexicography/attachments/20060531/efeca810/attachment.htm>


More information about the Lexicography mailing list